Precipitation prediction method, device and computer equipment

Abstract

The present application relates to a precipitation prediction method, device and computer equipment. The method comprises: determining the precipitation seasonal characteristic prediction result of the target area within a preset prediction period according to the historical data of each forecasting factor corresponding to the target precipitation time coefficient of the target area and the precipitation seasonal characteristic prediction model established in advance; the precipitation seasonal characteristic prediction result includes multiple precipitation seasonal characteristic prediction samples; determining the precipitation event simulation result of the target area within the prediction period according to the pre-constructed precipitation event simulation generation model and the precipitation historical data of the target area; the precipitation event simulation result includes multiple precipitation event simulation sub-results; determining the precipitation prediction result of the target area within the prediction period according to the precipitation seasonal characteristic prediction result and the precipitation event simulation result; the precipitation prediction result includes multiple precipitation prediction sub-results. The reliability of precipitation prediction can be improved by using the present method.

Description

Technical Field

[0001] The present application relates to the field of meteorological and hydrological technology, and in particular to a precipitation prediction method, apparatus, computer equipment, computer-readable storage medium, and computer program product. Background Art

[0002] Climate change has led to an acceleration of the atmospheric hydrological cycle and an increase in the variability of precipitation events, and the threat of global flood and drought disasters has become increasingly severe. Floods and droughts caused by extreme meteorological events can lead to damage to infrastructure, reduced crop yields, and other serious losses of people's lives, health, and property. They are one of the main natural disasters that threaten the high-quality development of the economy and society. In this context, scientifically estimating the occurrence characteristics of precipitation and its potential disaster risks is of great significance for flood and drought prevention.

[0003] The relevant precipitation prediction method is statistical forecasting, which mainly uses the teleconnection relationship between regional precipitation and different climate modes as the prediction basis to predict the precipitation prediction results in a certain area. However, the factors affecting the formation of precipitation events and precipitation intensity are very complex. Therefore, the prediction results based solely on teleconnection relationships often do not match the actual situation, resulting in poor reliability of precipitation prediction methods. Summary of the invention

[0004] Based on this, it is necessary to provide a precipitation prediction method, device, computer equipment, computer-readable storage medium and computer program product that can improve the reliability of precipitation prediction in response to the above technical problems.

[0005] In a first aspect, the present application provides a precipitation prediction method, comprising:

[0006] Obtain historical data of each forecast factor corresponding to the target precipitation time coefficient of the target area;

[0007] Determine the precipitation seasonal characteristic prediction result of the target area within a preset prediction period according to the pre-established precipitation seasonal characteristic prediction model and the historical data of each prediction factor; the precipitation seasonal characteristic prediction result includes a plurality of precipitation seasonal characteristic prediction samples;

[0008] Determine a precipitation event simulation result of the target area within the forecast period according to a pre-constructed precipitation event simulation generation model and the precipitation history data of the target area; the precipitation event simulation result includes a plurality of precipitation event simulation sub-results;

[0009] The precipitation prediction result of the target area within the prediction period is determined according to the precipitation season characteristic prediction result and the precipitation event simulation result; the precipitation prediction result includes a plurality of precipitation prediction sub-results.

[0010] In one embodiment, the precipitation seasonal characteristic prediction model includes a target precipitation time coefficient prediction sub-model corresponding to each target precipitation time coefficient, and the precipitation seasonal characteristic prediction result of the target area within a preset prediction period is determined according to the pre-established precipitation seasonal characteristic prediction model and the historical data of each prediction factor, including:

[0011] For each target precipitation time coefficient, historical data of each forecast factor corresponding to the target precipitation time coefficient is input into the target precipitation time coefficient prediction sub-model corresponding to the target precipitation time coefficient to obtain a probability distribution prediction result of the target precipitation time coefficient;

[0012] Sampling the probability distribution prediction results of each target precipitation time coefficient;

[0013] The samples of the target precipitation time coefficients obtained by sampling are multiplied and superimposed with the three-dimensional matrix of the target precipitation mode of the target area to obtain the precipitation seasonal characteristic prediction samples corresponding to this sampling;

[0014] The precipitation seasonal characteristic prediction samples corresponding to each sampling are used to form the precipitation seasonal characteristic prediction results of the target area within a preset prediction period.

[0015] In one embodiment, the precipitation event simulation generation model includes a precipitation event occurrence simulation model and a precipitation amount simulation model, and determining the precipitation event simulation result of the target area within the forecast period according to the pre-constructed precipitation event simulation generation model and the precipitation history data of the target area includes:

[0016] Determining a precipitation event occurrence simulation result according to the precipitation event occurrence simulation model and the precipitation historical data of the target area;

[0017] Determining a precipitation event simulation sub-result according to the precipitation simulation model, the precipitation event occurrence simulation result and the precipitation historical data of the target area;

[0018] Return to the step of determining the simulation result of the precipitation event occurrence in the target area within the forecast period according to the precipitation event occurrence simulation model and the precipitation historical data of the target area, until the preset precipitation event simulation stop condition is met, and the simulation sub-results of each precipitation event constitute the precipitation event simulation result of the target area within the forecast period.

[0019] In one embodiment, determining the precipitation prediction result of the target area within the prediction period according to the precipitation seasonal characteristic prediction result and the precipitation event simulation result includes:

[0020] Sampling with replacement is performed on both the precipitation seasonal characteristic prediction results and the precipitation event simulation results;

[0021] The L-BFGS-B quasi-Newton optimization algorithm is used to iteratively adjust the adjustment coefficients corresponding to the sub-results of the precipitation event simulation obtained in this sampling;

[0022] Performing a sliding average of a preset window length on the precipitation event simulation sub-result after each iterative adjustment to obtain a precipitation seasonal characteristic simulation sample after each iterative adjustment;

[0023] When the mean square error between the iteratively adjusted precipitation seasonal characteristic simulation samples and the precipitation seasonal characteristic prediction samples obtained by this sampling converges to the minimum, the precipitation event simulation sub-result after this iterative adjustment is used as the precipitation prediction sub-result corresponding to this sampling;

[0024] When the preset sampling stop condition is met, each of the precipitation prediction sub-results constitutes the precipitation prediction result of the target area within the prediction period.

[0025] In one embodiment, the method further comprises:

[0026] For each grid in the target area, the historical precipitation data of the grid is divided into a plurality of historical precipitation data samples according to the duration of the preset prediction period, and a three-dimensional matrix of historical precipitation data corresponding to each of the historical precipitation data samples is generated; the coordinate axes of the three-dimensional matrix of historical precipitation data are time, longitude and latitude respectively;

[0027] Performing a sliding average of a preset window length on the time coordinates of each three-dimensional matrix of historical precipitation data to obtain seasonal variation characteristic data;

[0028] Performing empirical orthogonal decomposition on the seasonal variation characteristic data to obtain the precipitation mode of the target area;

[0029] The Cattell scree test is used for the precipitation mode to determine the three-dimensional matrix of the target precipitation mode and the target precipitation time coefficient.

[0030] In one embodiment, the method further comprises:

[0031] Acquiring historical climate variable data of each preset climate variable of the target area;

[0032] According to the unit root test algorithm and the historical climate variable data, test whether the target precipitation time coefficient and the climate variable meet the stability condition;

[0033] When the target precipitation time coefficient and the climate variable meet the stationarity condition, for each target precipitation time coefficient, the historical climate variable data of the climate variable of the grid that passes the significance test is added to the climate variable set of the target precipitation time coefficient;

[0034] A principal component analysis method is applied to the climate variable set to obtain a data dimension reduction result;

[0035] The Cattell scree test is used on the data dimension reduction results to determine the prediction factor of the teleconnection corresponding to the target precipitation time coefficient.

[0036] In a second aspect, the present application also provides a precipitation prediction device, comprising:

[0037] The first acquisition module is used to acquire the historical data of each prediction factor corresponding to the target precipitation time coefficient of the target area;

[0038] A first determination module is used to determine the precipitation seasonal characteristic prediction result of the target area within a preset prediction period according to a pre-established precipitation seasonal characteristic prediction model and the historical data of each prediction factor; the precipitation seasonal characteristic prediction result includes a plurality of precipitation seasonal characteristic prediction samples;

[0039] A second determination module is used to determine a precipitation event simulation result of the target area within the forecast period according to a pre-constructed precipitation event simulation generation model and precipitation historical data of the target area; the precipitation event simulation result includes a plurality of precipitation event simulation sub-results;

[0040] The third determination module is used to determine the precipitation prediction result of the target area within the prediction period according to the precipitation season characteristic prediction result and the precipitation event simulation result; the precipitation prediction result includes multiple precipitation prediction sub-results.

[0041] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps described in the first aspect when executing the computer program.

[0042] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps described in the first aspect above.

[0043] In a fifth aspect, the present application further provides a computer program product, comprising a computer program, which implements the steps described in the first aspect when executed by a processor.

[0044] The above-mentioned precipitation prediction method, device, computer equipment, computer-readable storage medium and computer program product obtain the historical data of each prediction factor corresponding to the target precipitation time coefficient of the target area; determine the precipitation seasonal characteristic prediction result of the target area within a preset prediction period based on the pre-established precipitation seasonal characteristic prediction model and the historical data of each prediction factor; the precipitation seasonal characteristic prediction result includes a plurality of precipitation seasonal characteristic prediction samples; determine the precipitation event simulation result of the target area within the prediction period based on the pre-constructed precipitation event simulation generation model and the precipitation historical data of the target area; the precipitation event simulation result includes a plurality of precipitation event simulation sub-results; determine the precipitation prediction result of the target area within the prediction period based on the precipitation seasonal characteristic prediction result and the precipitation event simulation result; the precipitation prediction result includes a plurality of precipitation prediction sub-results. In this way, according to the pre-established precipitation seasonal characteristic prediction model and the historical data of each forecasting factor corresponding to the target precipitation time coefficient, multiple precipitation seasonal characteristic prediction samples of the target area within the prediction period are determined, and according to the pre-constructed precipitation event simulation generation model and precipitation historical data, multiple precipitation event simulation sub-results of the target area within the prediction period are determined, and then according to the precipitation seasonal characteristic prediction samples, the precipitation event simulation sub-results are adjusted and optimized to obtain multiple precipitation prediction sub-results of the target area within the prediction period, and the precipitation prediction is divided into a seasonal scale predictable part and an event scale uncertain part, so as to achieve accurate prediction of seasonal characteristics, and simulate different event occurrence scenarios based on the prediction results to achieve estimation of the uncertain part, taking into account the influence of uncertainty factors, and considering the complexity of the influencing factors of precipitation event formation and precipitation intensity, the reliability of precipitation prediction can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 A schematic diagram of a precipitation prediction method in one embodiment;

[0047] Figure 2 A schematic flow chart of a step of determining a precipitation seasonal characteristic prediction result of a target area within a preset prediction period according to a pre-established precipitation seasonal characteristic prediction model and historical data of each prediction factor in an embodiment;

[0048] Figure 3A flowchart of a step of determining a precipitation event simulation result of a target area within a forecast period according to a pre-built precipitation event simulation generation model and precipitation history data of the target area in an embodiment;

[0049] Figure 4 A flowchart of a step of determining a precipitation prediction result of a target area within a prediction period according to a precipitation season characteristic prediction result and a precipitation event simulation result in an embodiment;

[0050] Figure 5 A flowchart of the steps further included in the precipitation prediction method in one embodiment;

[0051] Figure 6 A flowchart of the steps further included in the precipitation prediction method in another embodiment;

[0052] Figure 7 A schematic diagram of the division of the A basin and its sub-basins in one embodiment;

[0053] Figure 8 It is a schematic diagram of the five-fold cross validation results of the prediction ability of each precipitation mode time coefficient in an embodiment;

[0054] Fig. 9 It is a schematic diagram of the prediction result of the seasonal characteristics of the average rainfall in the A basin in one embodiment;

[0055] Fig.10 A schematic diagram of the prediction results of the average annual rainfall of a surface according to a model in one embodiment;

[0056] Fig.11 A schematic diagram showing a comparison between the cumulative distribution function of daily precipitation in the A basin simulated by the model and the actual observation in one embodiment;

[0057] Fig.12 is a structural block diagram of a precipitation prediction device in one embodiment;

[0058] Fig.13 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0060] In one embodiment, Figure 1As shown, a precipitation prediction method is provided. This embodiment uses the method applied to a terminal as an example for illustration. It can be understood that the method can also be applied to a server, and can also be applied to a system including a terminal and a server, and is implemented through the interaction between the terminal and the server. Among them, the terminal can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, projection devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The head-mounted device can be a virtual reality (VR) device, an augmented reality (AR) device, smart glasses, etc. The server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud computing services. In this embodiment, the method includes the following steps:

[0061] Step 101, obtaining historical data of each forecasting factor corresponding to the target precipitation time coefficient of the target area.

[0062] In an embodiment of the present application, based on the previous hydrological and meteorological elements, the scientific prediction of the hydrological elements for a long period of time in the future (ten days to one year) is called a medium- and long-term hydrological forecast using the method of cause analysis and mathematical statistics. Medium- and long-term precipitation prediction is to predict the precipitation in the medium and long term in the future. The target area is the area where precipitation prediction is to be performed, and may include at least one of a river basin and a non-river basin. The target precipitation time coefficient is the time coefficient of the target precipitation mode, which is used to characterize the weight change of the corresponding spatial mode over time, and corresponds one to one with the target precipitation mode. A target precipitation time coefficient corresponds to at least one forecast factor, and these forecast factors are remotely correlated with the target precipitation time coefficient. The forecast factor can be a climate variable of a grid in the target area. The target area includes at least one grid. Climate variables are variables related to climate change.

[0063] Step 102, based on the pre-established precipitation seasonal characteristic prediction model and the historical data of each prediction factor, the precipitation seasonal characteristic prediction result of the target area within the preset prediction period is determined.

[0064] Among them, the precipitation seasonal characteristic prediction results include multiple precipitation seasonal characteristic prediction samples.

[0065] In an embodiment of the present application, the precipitation seasonal characteristic prediction model is used to predict the precipitation seasonal characteristics of the target area within the prediction period. The precipitation seasonal characteristic prediction result is used to characterize the precipitation seasonal characteristics of the target area within the prediction period. The prediction period is a future time period, which can be any time period from the next three days to the next year. For example, the prediction period can be the next year. The precipitation seasonal characteristic prediction model can include a target precipitation time coefficient prediction sub-model corresponding to each target precipitation time coefficient.

[0066] For each target precipitation time coefficient, the terminal inputs the historical data of each forecast factor corresponding to the target precipitation time coefficient into the target precipitation time coefficient prediction sub-model corresponding to the target precipitation time coefficient to obtain the probability distribution prediction result of the target precipitation time coefficient. Then, the terminal determines the precipitation season characteristic prediction sample according to the probability distribution prediction result of each target precipitation time coefficient. Then, the terminal uses each precipitation season characteristic prediction sample to form the precipitation season characteristic prediction result of the target area within the preset prediction period.

[0067] Step 103, determining the precipitation event simulation result of the target area within the forecast period according to the pre-built precipitation event simulation generation model and the precipitation history data of the target area.

[0068] Among them, the precipitation event simulation results include multiple precipitation event simulation sub-results.

[0069] In the embodiment of the present application, the precipitation event simulation generation model is used to simulate the occurrence of precipitation events and precipitation, and can be a weather generator. The weather generator can simulate and generate random precipitation sequences of any length. The frequency, intensity and spatiotemporal correlation of the generated precipitation sequences are consistent with the long-term statistical characteristics of the observed data. Different precipitation event scenarios can be simulated as meteorological inputs for flood and drought disaster assessment and impact analysis, but it cannot reflect the annual to decadal variability of precipitation.

[0070] The terminal determines the precipitation event simulation sub-result according to the precipitation event occurrence simulation model and the precipitation historical data of the target area. Then, the terminal repeatedly performs the step of determining the precipitation event simulation sub-result according to the precipitation event occurrence simulation model and the precipitation historical data of the target area until the preset precipitation event simulation stop condition is met. Then, the terminal uses each precipitation event simulation sub-result to form a precipitation event simulation result for the target area within the forecast period.

[0071] Step 104, determining the precipitation prediction result of the target area within the prediction period according to the precipitation season characteristic prediction result and the precipitation event simulation result.

[0072] Among them, the precipitation prediction result includes multiple precipitation prediction sub-results.

[0073] In an embodiment of the present application, the terminal adjusts the precipitation event simulation result according to the precipitation season characteristic prediction result, and uses the adjusted precipitation event simulation result as the precipitation prediction result for the target area within the prediction period.

[0074] In one example, the terminal adjusts the precipitation event simulation sub-result according to the precipitation seasonal characteristic prediction sample. Then, the terminal uses the precipitation event simulation sub-result that meets the preset first adjustment stop condition as the precipitation prediction sub-result. Then, the terminal returns to the step of adjusting the precipitation event simulation sub-result according to the precipitation seasonal characteristic prediction sample until the preset second adjustment stop condition is met. Then, the terminal uses each precipitation prediction sub-result to form a precipitation prediction result for the target area within the prediction period.

[0075] In the above precipitation prediction method, according to the historical data of each prediction factor corresponding to the pre-established precipitation seasonal characteristic prediction model and the target precipitation time coefficient, multiple precipitation seasonal characteristic prediction samples of the target area within the prediction period are determined, according to the pre-constructed precipitation event simulation generation model and precipitation historical data, multiple precipitation event simulation sub-results of the target area within the prediction period are determined, and then according to the precipitation seasonal characteristic prediction samples, the precipitation event simulation sub-results are adjusted and optimized to obtain multiple precipitation prediction sub-results of the target area within the prediction period, and the precipitation prediction is divided into a seasonal scale predictable part and an event scale uncertain part to achieve accurate prediction of seasonal characteristics, and simulate different event occurrence scenarios based on the prediction results to achieve estimation of the uncertain part, considering the influence of uncertainty factors, considering the complexity of the factors affecting the formation of precipitation events and precipitation intensity, can improve the reliability of precipitation prediction. In addition, this method also improves the statistical forecasting method to more accurately forecast the seasonal characteristics of precipitation, characterizes the uncertainty of precipitation events based on the weather generator concept, and couples the two to simulate various scenarios of precipitation that may occur in the next year. It can achieve accurate prediction of seasonal characteristics of precipitation at the basin scale and comprehensive simulation of different scenarios of precipitation events, solve the problem of insufficient reliability and difficulty in quantifying uncertainty in medium- and long-term precipitation forecasts, and provide a reliable reference for further assessment and prevention of water and drought disaster risks in the basin. Moreover, this method can achieve interannual precipitation prediction, which is more in line with actual needs and is of great significance to improving the basin's meteorological and hydrological forecasting and disaster prevention and mitigation capabilities, and protecting people's lives and property safety.

[0076] In an exemplary embodiment, Figure 2 As shown, the precipitation seasonal characteristic prediction model includes a target precipitation time coefficient prediction sub-model corresponding to each target precipitation time coefficient. According to the pre-established precipitation seasonal characteristic prediction model and the historical data of each prediction factor, the specific process of determining the precipitation seasonal characteristic prediction result of the target area within the preset prediction period includes the following steps:

[0077] Step 201, for each target precipitation time coefficient, input the historical data of each forecast factor corresponding to the target precipitation time coefficient into the target precipitation time coefficient prediction sub-model corresponding to the target precipitation time coefficient to obtain the probability distribution prediction result of the target precipitation time coefficient.

[0078] In an embodiment of the present application, the target precipitation time coefficient corresponds to the target precipitation time coefficient prediction submodel one by one. The target precipitation time coefficient prediction submodel can be a Bayesian Ridge Regression model. In this way, considering the relatively small number of samples for medium- and long-term precipitation prediction and the possible collinearity of climate variables, the use of the Bayesian Ridge Regression model can reduce the risk of model overfitting. The probability distribution prediction result includes the predicted values ​​of multiple target precipitation time coefficients.

[0079] In one example, the terminal can establish a Bayesian Ridge regression model for the target time coefficient sequence of each target precipitation mode. If differences are involved, the results are restored to the same order after regression. For each regression model, it is assumed that the regression coefficient vector w obeys the Gaussian distribution, and the regularization coefficients α and λ obey the gamma distribution. The posterior estimate of the parameters is obtained by maximizing the log marginal likelihood estimation during the fitting process. Since the regression coefficient obeys the normal distribution, the forecast result for each year is expressed as a linear combination of normally distributed random variables, which also obeys the normal distribution. The Bayesian Ridge regression model uses expectation and variance to characterize the forecast results of each target precipitation time coefficient each year.

[0080] Step 202: sampling the probability distribution prediction results of each target precipitation time coefficient.

[0081] In the embodiment of the present application, for each sampling, the terminal samples the probability distribution prediction results of each target precipitation time coefficient with replacement, and obtains the sample of each target precipitation time coefficient corresponding to this sampling. Among them, the sample of the target precipitation time coefficient is a predicted value of the target precipitation time coefficient included in the probability distribution prediction result of the target precipitation time coefficient.

[0082] Step 203, multiplying and superimposing the samples of each target precipitation time coefficient obtained by sampling with the three-dimensional matrix of the target precipitation mode of the target area to obtain the precipitation seasonal characteristic prediction sample corresponding to this sampling.

[0083] In the embodiment of the present application, for each target precipitation time coefficient, the terminal multiplies the sample of the target precipitation time coefficient obtained by this sampling with the three-dimensional matrix of the target precipitation mode corresponding to the target precipitation time coefficient to obtain the modal prediction result corresponding to the target precipitation time coefficient. Then, the terminal superimposes the modal prediction results corresponding to each target precipitation time coefficient to obtain the precipitation season characteristic prediction sample corresponding to this sampling.

[0084] Step 204, the precipitation seasonal characteristic prediction samples corresponding to each sampling are used to form the precipitation seasonal characteristic prediction result of the target area within the preset prediction period.

[0085] In the above precipitation prediction method, a corresponding prediction model is established for each target precipitation time coefficient, and the probability distribution of the target precipitation time coefficient is predicted through the model and historical data, and the probability distribution of each target precipitation time coefficient is sampled with replacement. After each sampling, the sample of each target precipitation time coefficient obtained by sampling is multiplied and superimposed with the three-dimensional matrix of the target precipitation mode of the target area to obtain the precipitation seasonal characteristic prediction sample, and each precipitation seasonal characteristic prediction sample constitutes the precipitation seasonal characteristic prediction result. In this way, the precipitation prediction is divided into a seasonal scale predictable part and an event scale uncertain part, and the accurate prediction of seasonal characteristics is achieved in advance. Considering the influence of deterministic factors, multiple precipitation seasonal characteristic prediction samples are generated through sampling with replacement, which are used as the long-term statistical characteristic basis for adjusting and optimizing the simulation results of precipitation events, which can improve the reliability of precipitation prediction.

[0086] In an exemplary embodiment, Figure 3 As shown, the precipitation event simulation generation model includes a precipitation event occurrence simulation model and a precipitation amount simulation model. According to the pre-built precipitation event simulation generation model and the precipitation historical data of the target area, the specific process of determining the precipitation event simulation result of the target area within the forecast period includes the following steps:

[0087] Step 301, determining a precipitation event occurrence simulation result according to a precipitation event occurrence simulation model and precipitation historical data of a target area.

[0088] In the embodiment of the present application, the precipitation event occurrence simulation model is used to simulate the occurrence of precipitation events. The precipitation event occurrence simulation result may be a precipitation event occurrence sequence.

[0089] The terminal binarizes the historical precipitation data of the target area to characterize the two states of sunny and rainy days. Specifically, if the daily rainfall is higher than 0.1 mm / day, the terminal determines that rainfall occurred on that day, which is a rainy day and is recorded as 1; otherwise, the terminal determines that no rainfall occurred on that day, which is a sunny day and is recorded as 0. Taking into account the differences in the probability of precipitation in different spatial locations and different months, the terminal divides and reorganizes the sunny and rainy sequences of each grid by month, and assumes that the sunny and rainy states of each grid in each month obey the first-order two-state Markov chain, and calculates the transition probability matrix separately. Then, the terminal uses a rank-related random number matrix to drive the Markov chain to simulate the spatial correlation of precipitation events between different grids, and obtains the simulation results of precipitation events.

[0090] Among them, the random number matrix with rank correlation is generated based on matrix decomposition and iterative optimization. The specific operation steps of this generation method include: the terminal calculates the Spearman rank correlation coefficient matrix (also known as the observation rank correlation coefficient matrix) of precipitation occurrence sequences in different grids according to the historical precipitation data of the target area, which is recorded as Then, the terminal Perform Cholesky decomposition to obtain the lower triangular matrix L, and multiply the Gaussian random number matrix X of size (n * k) by the transpose of L to obtain X * Where k is the number of grids, n is the length of the planned simulated precipitation sequence, and in an ideal case X * The rank correlation matrix will be equal to Then, the terminal applies the cumulative distribution function to convert X * It is converted into a random number matrix A that obeys a uniform distribution, and A is used as the probability input of the Markov chain state transition to obtain a simulated precipitation occurrence sequence. Specifically, if the uniformly distributed random number is greater than the critical probability of rainfall, the next day will be a rainy day, otherwise it will be sunny. The simulated rank correlation matrix of the precipitation occurrence sequence induced by the above method is Generally less than , that is, the simulated precipitation sequence is less correlated than the observation. The terminal can solve this problem through iterative optimization. The iterative equation is as follows:

[0091]

[0092] Where i is the iteration step, is the rank correlation coefficient matrix for Cholesky decomposition in the i-th iteration, is the correlation coefficient matrix actually calculated for the ith iteration of the simulated precipitation occurrence sequence, It is a parameter that determines the convergence speed and accuracy. The rank correlation matrix of the simulated precipitation occurrence sequence will continue to approach the actual observation, and the matrix 1-norm is used to measure the deviation. Under ideal conditions, the deviation can be less than 1%. When the preset rank correlation optimization stop condition is met, the terminal uses the currently optimized simulated precipitation occurrence sequence as the simulation result of the precipitation event. The rank correlation optimization stop condition can be that the matrix 1-norm is less than a preset deviation threshold and / or reaches a preset rank correlation optimization stop number threshold.

[0093] Step 302, determining a precipitation event simulation sub-result according to the precipitation simulation model, the precipitation event occurrence simulation result and the precipitation history data of the target area.

[0094] In the embodiment of the present application, the precipitation event simulation sub-result is used to characterize the occurrence and amount of precipitation.

[0095] After completing the simulation of the precipitation sequence, the terminal generates precipitation for each grid where precipitation events occur. Specifically, the terminal describes the correlation between precipitation events and spatial range through conditional probability, and defines the occurrence index K, where O is the occurrence state vector, recording whether precipitation occurs in grids other than the grid where it is located, and U is the unit vector. When precipitation occurs at all stations, O is equal to U. is a correlation vector, which records the correlation degree between the grid and other grids, and can be expressed as:

[0096]

[0097] The terminal calculates the occurrence index K for each grid where a precipitation event occurs, which is used to characterize the spatial range of the precipitation event. Specifically, the terminal evenly divides all K values ​​of each grid into four intervals, and divides the precipitation observed in each grid into four categories according to the interval to which the occurrence index belongs to establish the correlation between precipitation and spatial range. The terminal fits each category of precipitation in each grid using the Kappa-3 parameter distribution, and calculates the K value of the simulated precipitation occurrence sequence. Then, the terminal randomly generates precipitation according to the fitting result of the Kappa-3 parameter distribution corresponding to the class to obtain the final precipitation event simulation sub-result.

[0098] In one example, the above method needs to be adjusted accordingly when applied to the simulation of precipitation events in watersheds of different sizes. For large watersheds, since the dimension of the rank correlation matrix C is equal to the number of grids, too many grids will lead to ill-conditioned problems in the correlation decomposition and calculation of the rank correlation matrix. The watershed can be divided into multiple 3*3 grids, and the central grids can be combined into a new matrix for precipitation event simulation, and then the results can be interpolated back to the 3*3 grid. For small watersheds, the correlation between grid precipitation sequences is generally strong, and precipitation events covering the entire watershed frequently occur in summer (K=1). For this reason, the rank correlation of the precipitation of this part of precipitation events needs to be adjusted again to ensure the simulation effect of extreme events.

[0099] Step 303, returns to the step of determining the simulation result of the precipitation event in the target area within the forecast period according to the precipitation event occurrence simulation model and the precipitation historical data of the target area, until the preset precipitation event simulation stop condition is met, and the simulation sub-results of each precipitation event constitute the precipitation event simulation result of the target area within the forecast period.

[0100] In the embodiment of the present application, the terminal returns the step of determining the simulation result of the occurrence of a precipitation event in the target area within the forecast period according to the precipitation event occurrence simulation model and the precipitation history data of the target area. Until the preset precipitation event simulation stop condition is met, the terminal will form the precipitation event simulation result of the target area within the forecast period with each precipitation event simulation sub-result.

[0101] In the above precipitation prediction method, the precipitation event occurrence simulation result is determined according to the precipitation event occurrence simulation model and the precipitation historical data of the target area; the precipitation event simulation sub-result is determined according to the precipitation amount simulation model, the precipitation event occurrence simulation result and the precipitation historical data of the target area; the step of determining the precipitation event occurrence simulation result of the target area within the prediction period according to the precipitation event occurrence simulation model and the precipitation historical data of the target area is returned until the preset precipitation event simulation stop condition is met, and each precipitation event simulation sub-result constitutes the precipitation event simulation result of the target area within the prediction period. In this way, the weather generator first simulates the generation of precipitation events, and then simulates the precipitation amount, and simulates and generates random precipitation sequences of any length. The frequency, intensity and spatiotemporal correlation of the generated precipitation sequences are consistent with the long-term statistical characteristics of the observed data, simulating different precipitation event occurrence scenarios, realizing the estimation of the uncertain part of precipitation prediction, considering the influence of uncertainty factors, and improving the reliability of precipitation prediction.

[0102] In an exemplary embodiment, Figure 4 As shown in FIG, based on the precipitation seasonal characteristic prediction results and the precipitation event simulation results, the specific process of determining the precipitation prediction results of the target area within the prediction period includes the following steps:

[0103] Step 401 , sampling with replacement is performed on both the precipitation seasonal characteristic prediction results and the precipitation event simulation results.

[0104] In the embodiment of the present application, the terminal performs sampling with replacement on the precipitation season characteristic prediction results and the precipitation event simulation results respectively.

[0105] Step 402, using the L-BFGS-B quasi-Newton adjustment algorithm, iteratively adjusts the adjustment coefficients corresponding to the precipitation event simulation sub-results obtained in this sampling.

[0106] In the embodiment of the present application, the adjustment coefficient may be an adjustment coefficient corresponding to a unit precipitation in the precipitation event simulation sub-result, and the adjustment coefficient is multiplied by the unit precipitation to achieve iterative adjustment of the precipitation event simulation sub-result. The unit precipitation may be a monthly precipitation. When the prediction period is 1 year, the adjustment coefficient is 12, and each monthly precipitation in the precipitation event simulation sub-result corresponds to an adjustment coefficient.

[0107] In one example, the terminal limits the size of each adjustment coefficient to the interval [0.5, 2]. This can preserve the variability characteristics of precipitation and prevent over-smoothing.

[0108] Step 403, performing a sliding average of a preset window length on the precipitation event simulation sub-result after each iteration adjustment to obtain a precipitation seasonal characteristic simulation sample after each iteration adjustment.

[0109] In the embodiment of the present application, the window length may be 90 days. The precipitation event simulation sub-result may include the daily precipitation distribution of each grid in a year, and the seasonal characteristics of the precipitation simulation may be obtained by applying a 90-day sliding average.

[0110] Step 404, when the mean square error between the iteratively adjusted precipitation seasonal characteristic simulation samples and the precipitation seasonal characteristic prediction samples obtained by this sampling converges to a minimum, the precipitation event simulation sub-result after this iterative adjustment is used as the precipitation prediction sub-result corresponding to this sampling.

[0111] In the embodiment of the present application, for each iterative adjustment, the terminal calculates the mean square error after the iterative adjustment based on the precipitation seasonal characteristic simulation sample after the iterative adjustment and the precipitation seasonal characteristic prediction sample obtained by this sampling. When the mean square error of the precipitation seasonal characteristic simulation sample after the iterative adjustment and the precipitation seasonal characteristic prediction sample obtained by this sampling converges to the minimum, the terminal uses the precipitation event simulation sub-result after the iterative adjustment as the precipitation prediction sub-result corresponding to this sampling. Among them, the number of iterative adjustments can be limited and can be less than the preset iterative adjustment threshold.

[0112] Step 405, when the preset sampling stop condition is met, each precipitation prediction sub-result is used to form a precipitation prediction result for the target area within the prediction period.

[0113] In an embodiment of the present application, the sampling stop condition may be that the number of samplings reaches a preset sampling number threshold.

[0114] In the above precipitation prediction method, the precipitation seasonal characteristic prediction results and the precipitation event simulation results are sampled with replacement multiple times, and after each sampling, the precipitation event simulation sub-results are iteratively adjusted according to the precipitation seasonal characteristic prediction samples, so that the seasonal characteristics of the simulation sub-results converge with the predicted seasonal characteristics, and precipitation prediction is achieved by combining climate variability signals. This solves the main problem that the precipitation sequence generated by the weather generator cannot reflect the annual to decadal variability of precipitation, and this part deviates from the long-term statistical characteristics. While considering the complexity of the factors affecting the formation of precipitation events and precipitation intensity, the reliability of precipitation prediction can be further improved. In addition, this method randomly extracts from the obtained basin precipitation seasonal characteristic prediction samples as the seasonal characteristic prediction results, calculates the mean square error of the seasonal characteristic simulation and seasonal characteristic prediction results for each grid, multiplies the precipitation of the precipitation event simulation by month, and uses the L-BFGS-B quasi-Newton optimization algorithm to minimize the mean square error of the seasonal characteristics of each grid in a finite number of iterations. By coupling the precipitation seasonal characteristic prediction and event simulation results, a precipitation scenario simulation under the seasonal characteristic prediction results can be obtained. Repeating the above operations in large numbers can obtain possible simulations of various precipitation scenarios for the next year, and the simulation results can be used for further hydrological analysis and disaster risk assessment.

[0115] In an exemplary embodiment, Figure 5 As shown, the method further comprises the following steps:

[0116] Step 501 , for each grid in the target area, according to the duration of the preset prediction period, the historical precipitation data of the grid is divided into a plurality of historical precipitation data samples, and a three-dimensional matrix of historical precipitation data corresponding to each historical precipitation data sample is generated.

[0117] Among them, the coordinate axes of the three-dimensional matrix of historical precipitation data are time, longitude and latitude.

[0118] In an embodiment of the present application, for each grid in the target area, the terminal divides the historical precipitation data of the grid into multiple historical precipitation data samples according to the duration of the preset prediction period, and generates a three-dimensional matrix of historical precipitation data corresponding to each historical precipitation data sample. The historical precipitation data may be grid data. When the prediction period is the next year, leap years are recorded as ordinary years, and the time coordinates of each historical precipitation data sample contain 365 days.

[0119] In one example, the terminal obtains historical precipitation data for each grid in the target area.

[0120] In another example, the terminal obtains historical observation data of each precipitation station in the target area, and then interpolates the historical observation data into grid data to obtain historical precipitation data of each grid in the target area.

[0121] Step 502 , performing a sliding average of a preset window length on the time coordinates of each three-dimensional matrix of historical precipitation data to obtain seasonal variation characteristic data.

[0122] In the embodiment of the present application, the window length may be 90 days. In this way, a sliding average with a window width of 90 days is performed on the time coordinate, and it is considered that the data after the sliding average represents the variation characteristics of the basin precipitation on a seasonal scale.

[0123] Step 503: Perform empirical orthogonal decomposition on the seasonal variation characteristic data to obtain the precipitation mode of the target area.

[0124] In an embodiment of the present application, empirical orthogonal decomposition (EOF), also known as empirical orthogonal function analysis method, or eigenvector analysis, is a method for analyzing structural features in matrix data and extracting main data feature quantities. In EOF, we pay more attention to the eigenvectors of the obtained covariance matrix, also called spatial eigenvectors (spatial modes). The i-th eigenvector is called the i-th mode: EOF_{i}. The spatial mode reflects the spatial distribution characteristics of the element field to a certain extent. If it is a calculation for meteorological data, it is a re-fitted element field or mode. PC (principal component) corresponds to time changes, also known as time coefficient, which reflects the weight change of the corresponding spatial mode over time. The precipitation mode is the mode of precipitation data. When the prediction period is the next year, the precipitation mode is also called the precipitation year mode.

[0125] Step 504: Cattell scree test is used on the precipitation mode to determine the three-dimensional matrix of the target precipitation mode and the target precipitation time coefficient.

[0126] In an embodiment of the present application, the terminal uses a Cattell gravel test on the precipitation mode, determines the retained target precipitation mode in the precipitation mode of the target area, and obtains a three-dimensional matrix of the target precipitation mode and a target precipitation time coefficient corresponding to the target precipitation mode. Among them, the three-dimensional matrix of the target precipitation mode is a three-dimensional spatiotemporal modal matrix. The target precipitation time coefficient can be in the form of a time coefficient sequence.

[0127] In the above precipitation prediction method, a three-dimensional matrix of historical precipitation data is generated, and a sliding average is performed on it to obtain seasonal variation characteristic data. The seasonal variation characteristic data is subjected to empirical orthogonal decomposition, precipitation mode is extracted, and a Cattell gravel test is performed on the precipitation mode. The target precipitation mode is screened to obtain the three-dimensional matrix of the target precipitation mode and its corresponding target precipitation time coefficient. When predicting the seasonal characteristics of precipitation, selecting accurate modes and time coefficients can improve the accuracy of the seasonal characteristics prediction of precipitation, thereby further improving the reliability of precipitation prediction.

[0128] In an exemplary embodiment, Figure 6 As shown, the method further comprises the following steps:

[0129] Step 601, obtaining historical climate variable data of preset climate variables in the target area.

[0130] In the embodiment of the present application, the climate variables may include at least one of the monthly average grid data of sea surface temperature, land surface temperature, 850 hPa geopotential height, 200 hPa geopotential height and national snow depth. The historical climate variable data may be multi-source observation or reanalysis data, and may be a time series of climate variables extracted monthly and changing year by year.

[0131] Step 602, based on the unit root test algorithm and the historical climate variable data, test whether the target precipitation time coefficient and the climate variable meet the stability condition.

[0132] In the embodiment of the present application, the unit root test algorithm may be ADF, which is a type of unit root test, and its full English name is Augmented Dickey-Fuller Tested, and its Chinese name is Extended Dickey-Fuller Test.

[0133] Step 603: When the target precipitation time coefficient and the climate variable meet the stationarity condition, for each target precipitation time coefficient, the historical climate variable data of the climate variable of the grid that passes the significance test is added to the climate variable set of the target precipitation time coefficient.

[0134] In an embodiment of the present application, when the target precipitation time coefficient and the climate variable meet the stationarity condition, for each target precipitation time coefficient, the historical climate variable data of each climate variable of each grid in the target area are respectively tested for significance. Specifically, the terminal applies the Granger causality test with a maximum lag of 3 years to analyze the causality between the target precipitation time coefficient and each climate variable of each grid in the target area at a significance level of α=0.05. Then, the terminal adds the historical climate variable data of the climate variables of the grid that passes the significance test to the climate variable set of the target precipitation time coefficient. Among them, one target precipitation time coefficient corresponds to a set of climate variables that have causality with it.

[0135] In one embodiment, if the target precipitation time coefficient and the climate variable do not meet the stationarity condition,

[0136] The terminal performs the same-order difference on the parameters that do not meet the stationary condition until the parameters meet the stationary condition.

[0137] Step 604, using principal component analysis on the climate variable set to obtain data dimensionality reduction results.

[0138] In the embodiments of the present application, principal component analysis (PCA), also known as principal component analysis or principal component regression analysis, aims to transform multiple indicators into a few comprehensive indicators by using the idea of ​​dimensionality reduction.

[0139] Step 605: Cattell scree test is used on the data dimension reduction result to determine the prediction factor of the teleconnection corresponding to the target precipitation time coefficient.

[0140] In the embodiment of the present application, the terminal uses Cattell scree test on the data dimension reduction result, and determines the prediction factor of the telecorrelation corresponding to the target precipitation time coefficient in the climate variable set of the target precipitation time coefficient.

[0141] In the above precipitation prediction method, when the target precipitation time coefficient and the climate variable meet the stationarity condition, the historical climate variable data of the climate variable of the grid that passes the significance test are added to the climate variable set of the target precipitation time coefficient, and then the climate variable set is subjected to principal component analysis and Cattell gravel test in turn to screen the accurate prediction factor of the target precipitation time coefficient, which can improve the accuracy of the prediction of the target precipitation time coefficient and the accuracy of the prediction of the precipitation seasonal characteristics, thereby further improving the reliability of precipitation prediction.

[0142] In one embodiment, the target area is the A basin. The A basin and its sub-basins are divided as follows: Figure 7The precipitation observation data used is derived from the daily precipitation grid dataset, which is obtained by interpolating and correcting the precipitation observation data of 2839 meteorological stations. The spatial resolution is 0.25°×0.25°, which can more realistically reflect the actual precipitation characteristics of various regions. The prediction model is trained and validated within the time range of 1991 to 2022.

[0143] According to the latitude and longitude range of the A basin, the grid data was clipped to the area of ​​96°E~119°E, 32°N~42°N, and the clipped data was sorted into 32 three-dimensional array samples by year, and the size of each sample was (365,44,96), corresponding to the time, latitude, and longitude coordinates, respectively. A 90-day sliding average was applied to the time axis of each sample to obtain the seasonal characteristics of precipitation for each year. The smoothed data was empirically orthogonal decomposition, and the Cattell gravel test was used to determine that six main precipitation annual modes (EOF) should be retained, which can explain about 60% of the variance. These six modes are considered to be the predictable part of rainfall one year in advance. After decomposition, six precipitation annual mode three-dimensional matrices and six corresponding time coefficient sequences (PC) with a length of 32 were obtained.

[0144] The ocean and land surface temperature data in the multi-source data used to construct the forecast factors come from HadCRUT5 analysis data, the 850hpa geopotential height and 200hpa geopotential height data come from ERA5 (ECMWF Reanalysis v5, the fifth generation of atmospheric reanalysis data sets for global climate by the European Centre for Medium-Range Weather Forecasts (ECMWF)) reanalysis data, and the snow depth data comes from the snow depth long series data set. All of the above data are monthly average grid data. Among them, HadCRUT5 and ERA5 data are updated in real time to meet the data timeliness requirements for one-year-ahead forecasts; the snow depth long series data set is currently updated to 2023, but the AMSR2 satellite remote sensing data as the original data source can be obtained in real time. According to the method introduced in the data set, the snow depth can be inverted and calculated to meet the forecast data requirements.

[0145] The ADF unit root test was applied and it was found that the time coefficients did not meet the stationary condition. This may be due to the periodicity of climate variability and the trend of precipitation growth caused by global warming. The first-order difference processing of each time coefficient and multi-source data met the stationary condition. The Granger causality test with a maximum lag of 3 years was used to analyze the causality of each time coefficient and each grid of multi-source data at a significance level of α=0.05. The time series of climate variables in the grid that passed the significance test were added to the large set of predictors corresponding to the time coefficients. Six time coefficients corresponded to six large sets of predictors. Principal component analysis was applied to each set for data dimensionality reduction. When 10 principal components were retained, more than 60% of the variance of each large set could be explained. Each time coefficient corresponded to 10 predictors.

[0146] A Bayesian ridge regression model was established for the time coefficient of each precipitation mode for prediction. A 5-fold cross-validation method was used to evaluate the performance and generalization ability of the model. A 10-year continuous test was conducted with 2013-2022 as the prediction period to further evaluate the model's predictive ability for the seasonal characteristics of basin precipitation. Figure 8 is the five-fold cross-validation result of the prediction ability of each precipitation mode time coefficient, and the determination coefficient R 2 It reaches about 0.8, and the prediction ability for the predictable part of precipitation in Basin A composed of the superposition of annual modes is strong. The difference between the prediction results and the actual observation results needs to be further verified. Fig. 9 This is the prediction result of the seasonal characteristics of the average rainfall in the A basin by this method, and the coefficient of determination is above 0.9; Fig.10 The model predicts the average annual rainfall in the area. The deviation between the prediction and observation is small and the actual observation is within the confidence interval of the prediction. In summary, this method can accurately predict the seasonal characteristics of precipitation in Basin A one year in advance.

[0147] Since the A basin is a large basin, the research scope contains more than 3000 grids. The basin is divided into several 3*3 grids and the central grid is reorganized into a smaller matrix. The size of the matrix is ​​(365, 15, 32), and the new matrix contains 480 grids. The precipitation event simulation generation model of the A basin is constructed according to the above method. The difference between the rank correlation matrix of the monthly precipitation occurrence sequence of the model and the observation ( ) has a 1-norm less than The 1-norm of the A-basin is 15%. Higher precision error requirements will lead to slower convergence speed. The generated simulated precipitation series is bilinearly interpolated back to the original size to obtain the simulated results of precipitation in Basin A. Fig.11 This is a schematic diagram comparing the cumulative distribution functions of daily precipitation in Basin A between the model simulation and actual observation. This method can better restore the probability distribution characteristics of daily precipitation in Basin A and its sub-basins.

[0148] The precipitation event simulation results give the daily precipitation of each grid in a year. The precipitation is adjusted monthly to make the precipitation seasonal characteristics of the simulation results consistent with the prediction. The specific operation is to introduce 12 adjustment coefficients for each grid corresponding to the 12 months of the year, multiply the daily precipitation of each month of the grid by the corresponding adjustment coefficient, change the size of the adjustment coefficient, and iterate and optimize until the seasonal characteristics of the precipitation sequence of each grid after adjustment converge to the minimum mean square error with the prediction results. In the actual operation process, the size interval of each adjustment coefficient is limited to [0.5, 2] to retain the variability characteristics of precipitation and prevent over-smoothing. This method has a strong simulation ability for the daily scale variability of basin precipitation.

[0149] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0150] Based on the same inventive concept, the embodiment of the present application also provides a precipitation prediction device for implementing the precipitation prediction method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more precipitation prediction device embodiments provided below can refer to the limitations of the precipitation prediction method above, and will not be repeated here.

[0151] In an exemplary embodiment, Fig.12 As shown, a precipitation prediction device 1200 is provided, comprising: a first acquisition module 1210, a first determination module 1220, a second determination module 1230 and a third determination module 1240, wherein:

[0152] The first acquisition module 1210 is used to acquire historical data of each prediction factor corresponding to the target precipitation time coefficient of the target area;

[0153] The first determination module 1220 is used to determine the precipitation seasonal characteristic prediction result of the target area within a preset prediction period according to the pre-established precipitation seasonal characteristic prediction model and the historical data of each prediction factor; the precipitation seasonal characteristic prediction result includes a plurality of precipitation seasonal characteristic prediction samples;

[0154] The second determination module 1230 is used to determine the precipitation event simulation result of the target area within the forecast period according to the pre-built precipitation event simulation generation model and the precipitation history data of the target area; the precipitation event simulation result includes a plurality of precipitation event simulation sub-results;

[0155] The third determination module 1240 is used to determine the precipitation prediction result of the target area within the prediction period according to the precipitation season characteristic prediction result and the precipitation event simulation result; the precipitation prediction result includes multiple precipitation prediction sub-results.

[0156] Optionally, the precipitation seasonal characteristic prediction model includes a target precipitation time coefficient prediction sub-model corresponding to each target precipitation time coefficient, and the first determination module 1220 is specifically used to:

[0157] For each target precipitation time coefficient, historical data of each forecast factor corresponding to the target precipitation time coefficient is input into the target precipitation time coefficient prediction sub-model corresponding to the target precipitation time coefficient to obtain a probability distribution prediction result of the target precipitation time coefficient;

[0158] Sampling the probability distribution prediction results of each target precipitation time coefficient;

[0159] The samples of the target precipitation time coefficients obtained by sampling are multiplied and superimposed with the three-dimensional matrix of the target precipitation mode of the target area to obtain the precipitation seasonal characteristic prediction samples corresponding to this sampling;

[0160] The precipitation seasonal characteristic prediction samples corresponding to each sampling are used to form the precipitation seasonal characteristic prediction results of the target area within a preset prediction period.

[0161] Optionally, the precipitation event simulation generation model includes a precipitation event occurrence simulation model and a precipitation amount simulation model, and the second determination module 1230 is specifically used to:

[0162] Determining a precipitation event occurrence simulation result according to the precipitation event occurrence simulation model and the precipitation historical data of the target area;

[0163] Determining a precipitation event simulation sub-result according to the precipitation simulation model, the precipitation event occurrence simulation result and the precipitation historical data of the target area;

[0164] Return to the step of determining the simulation result of the precipitation event occurrence in the target area within the forecast period according to the precipitation event occurrence simulation model and the precipitation historical data of the target area, until the preset precipitation event simulation stop condition is met, and the simulation sub-results of each precipitation event constitute the precipitation event simulation result of the target area within the forecast period.

[0165] Optionally, the third determining module 1240 is specifically configured to:

[0166] Sampling with replacement is performed on both the precipitation seasonal characteristic prediction results and the precipitation event simulation results;

[0167] The L-BFGS-B quasi-Newton optimization algorithm is used to iteratively adjust the adjustment coefficients corresponding to the sub-results of the precipitation event simulation obtained in this sampling;

[0168] Performing a sliding average of a preset window length on the precipitation event simulation sub-result after each iteration adjustment to obtain a precipitation seasonal characteristic simulation sample after each iteration adjustment;

[0169] When the mean square error between the iteratively adjusted precipitation seasonal characteristic simulation samples and the precipitation seasonal characteristic prediction samples obtained by this sampling converges to the minimum, the precipitation event simulation sub-result after this iterative adjustment is used as the precipitation prediction sub-result corresponding to this sampling;

[0170] When the preset sampling stop condition is met, each of the precipitation prediction sub-results constitutes the precipitation prediction result of the target area within the prediction period.

[0171] Optionally, the device 1200 further includes:

[0172] A generation module is used to divide the historical precipitation data of each grid in the target area into multiple historical precipitation data samples according to the duration of a preset prediction period, and generate a three-dimensional matrix of historical precipitation data corresponding to each of the historical precipitation data samples; the coordinate axes of the three-dimensional matrix of historical precipitation data are time, longitude and latitude respectively;

[0173] A sliding average module is used to perform sliding average of a preset window length on the time coordinates of each three-dimensional matrix of historical precipitation data to obtain seasonal variation characteristic data;

[0174] An empirical orthogonal decomposition module, used to perform empirical orthogonal decomposition on the seasonal variation characteristic data to obtain the precipitation mode of the target area;

[0175] The first scree test module is used to use Cattell scree test on the precipitation mode to determine the three-dimensional matrix of the target precipitation mode and the target precipitation time coefficient.

[0176] Optionally, the device 1200 further includes:

[0177] A second acquisition module is used to acquire historical climate variable data of each preset climate variable in the target area;

[0178] A first testing module is used to test whether the target precipitation time coefficient and the climate variable meet the stability condition according to the unit root test algorithm and the historical climate variable data;

[0179] A second testing module is used for, when the target precipitation time coefficient and the climate variable meet the stationarity condition, for each target precipitation time coefficient, adding the historical climate variable data of the climate variable of the grid that passes the significance test to the climate variable set of the target precipitation time coefficient;

[0180] A principal component analysis module, used for applying principal component analysis to the climate variable set to obtain a data dimension reduction result;

[0181] The second scree test module is used to use Cattell scree test on the data dimension reduction result to determine the prediction factor of the teleconnection corresponding to the target precipitation time coefficient.

[0182] Each module in the above precipitation prediction device can be implemented in whole or in part by software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0183] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Fig.13 As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, near field communication (NFC) or other technologies. When the computer program is executed by the processor, a precipitation prediction method is implemented. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0184] Those skilled in the art will understand that Fig.13 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0185] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0186] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0187] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0188] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0189] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0190] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0191] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be construed as limiting the scope of the present application. It should be noted that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A precipitation prediction method, characterized in that: The method comprises: Obtain historical data of each forecast factor corresponding to the target precipitation time coefficient of the target area; Determine the precipitation seasonal characteristic prediction result of the target area within a preset prediction period according to the pre-established precipitation seasonal characteristic prediction model and the historical data of each prediction factor; the precipitation seasonal characteristic prediction result includes a plurality of precipitation seasonal characteristic prediction samples; Determine a precipitation event simulation result of the target area within the forecast period according to a pre-constructed precipitation event simulation generation model and the precipitation history data of the target area; the precipitation event simulation result includes a plurality of precipitation event simulation sub-results; The precipitation prediction result of the target area within the prediction period is determined according to the precipitation season characteristic prediction result and the precipitation event simulation result; the precipitation prediction result includes a plurality of precipitation prediction sub-results.

2. The method according to claim 1, characterized in that The precipitation seasonal characteristic prediction model includes a target precipitation time coefficient prediction sub-model corresponding to each target precipitation time coefficient. The precipitation seasonal characteristic prediction result of the target area within a preset prediction period is determined based on the pre-established precipitation seasonal characteristic prediction model and the historical data of each prediction factor, including: For each target precipitation time coefficient, historical data of each forecast factor corresponding to the target precipitation time coefficient is input into the target precipitation time coefficient prediction sub-model corresponding to the target precipitation time coefficient to obtain a probability distribution prediction result of the target precipitation time coefficient; Sampling the probability distribution prediction results of each target precipitation time coefficient; The samples of the target precipitation time coefficients obtained by sampling are multiplied and superimposed with the three-dimensional matrix of the target precipitation mode of the target area to obtain the precipitation seasonal characteristic prediction samples corresponding to this sampling; The precipitation seasonal characteristic prediction samples corresponding to each sampling are used to form the precipitation seasonal characteristic prediction results of the target area within a preset prediction period.

3. The method according to claim 1, characterized in that: The precipitation event simulation generation model includes a precipitation event occurrence simulation model and a precipitation amount simulation model. The step of determining the precipitation event simulation result of the target area within the prediction period according to the pre-constructed precipitation event simulation generation model and the precipitation history data of the target area includes: Determining a precipitation event occurrence simulation result according to the precipitation event occurrence simulation model and the precipitation historical data of the target area; Determining a precipitation event simulation sub-result according to the precipitation simulation model, the precipitation event occurrence simulation result and the precipitation historical data of the target area; Return to the step of determining the simulation result of the precipitation event occurrence in the target area within the forecast period according to the precipitation event occurrence simulation model and the precipitation historical data of the target area, until the preset precipitation event simulation stop condition is met, and the simulation sub-results of each precipitation event constitute the precipitation event simulation result of the target area within the forecast period.

4. The method according to claim 1, characterized in that The step of determining the precipitation prediction result of the target area within the prediction period according to the precipitation seasonal characteristic prediction result and the precipitation event simulation result includes: Sampling with replacement is performed on both the precipitation seasonal characteristic prediction results and the precipitation event simulation results; The L-BFGS-B quasi-Newton optimization algorithm is used to iteratively adjust the adjustment coefficients corresponding to the sub-results of the precipitation event simulation obtained in this sampling; Performing a sliding average of a preset window length on the precipitation event simulation sub-result after each iteration adjustment to obtain a precipitation seasonal characteristic simulation sample after each iteration adjustment; When the mean square error between the iteratively adjusted precipitation seasonal characteristic simulation samples and the precipitation seasonal characteristic prediction samples obtained by this sampling converges to the minimum, the precipitation event simulation sub-result after this iterative adjustment is used as the precipitation prediction sub-result corresponding to this sampling; When the preset sampling stop condition is met, each of the precipitation prediction sub-results constitutes the precipitation prediction result of the target area within the prediction period.

5. The method according to claim 2, characterized in that: The method further comprises: For each grid in the target area, the historical precipitation data of the grid is divided into a plurality of historical precipitation data samples according to the duration of the preset prediction period, and a three-dimensional matrix of historical precipitation data corresponding to each of the historical precipitation data samples is generated; the coordinate axes of the three-dimensional matrix of historical precipitation data are time, longitude and latitude respectively; Performing a sliding average of a preset window length on the time coordinates of each three-dimensional matrix of historical precipitation data to obtain seasonal variation characteristic data; Performing empirical orthogonal decomposition on the seasonal variation characteristic data to obtain the precipitation mode of the target area; The Cattell scree test is used for the precipitation mode to determine the three-dimensional matrix of the target precipitation mode and the target precipitation time coefficient.

6. The method according to claim 1, characterized in that The method further comprises: Acquiring historical climate variable data of each preset climate variable of the target area; According to the unit root test algorithm and the historical climate variable data, test whether the target precipitation time coefficient and the climate variable meet the stability condition; When the target precipitation time coefficient and the climate variable meet the stationarity condition, for each target precipitation time coefficient, the historical climate variable data of the climate variable of the grid that passes the significance test is added to the climate variable set of the target precipitation time coefficient; A principal component analysis method is applied to the climate variable set to obtain a data dimension reduction result; The Cattell scree test is used on the data dimension reduction results to determine the prediction factor of the teleconnection corresponding to the target precipitation time coefficient.

7. A precipitation prediction device, characterized in that: The device comprises: The first acquisition module is used to acquire the historical data of each prediction factor corresponding to the target precipitation time coefficient of the target area; A first determination module is used to determine the precipitation seasonal characteristic prediction result of the target area within a preset prediction period according to a pre-established precipitation seasonal characteristic prediction model and the historical data of each prediction factor; the precipitation seasonal characteristic prediction result includes a plurality of precipitation seasonal characteristic prediction samples; A second determination module is used to determine a precipitation event simulation result of the target area within the forecast period according to a pre-constructed precipitation event simulation generation model and precipitation historical data of the target area; the precipitation event simulation result includes a plurality of precipitation event simulation sub-results; The third determination module is used to determine the precipitation prediction result of the target area within the prediction period according to the precipitation season characteristic prediction result and the precipitation event simulation result; the precipitation prediction result includes multiple precipitation prediction sub-results.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images