Lightweight short-term precipitation forecasting method and device considering time and parameter correlation

Abstract

The present invention discloses a lightweight short-term precipitation forecasting method and device considering time and parameter correlation. The method includes: obtaining precipitation and precipitation-related influencing factors in a target measurement area; inputting the precipitation and precipitation-related influencing factors into a trained time series forecasting model to output first precipitation forecasting data for a future preset time period; inputting the precipitation-related influencing factors into the trained time series forecasting model to output key precipitation influencing factors for the future preset time period, and inputting the key precipitation influencing factors for the future preset time period into a trained relationship model to output second precipitation forecasting data for the future preset time period; and obtaining a final precipitation forecasting result through weighted fitting of the first precipitation forecasting data and the second precipitation forecasting data. Based on Beidou meteorological monitoring data and an artificial intelligence model, the present invention considers time and parameter correlation, realizes a more accurate expression of short-term precipitation conditions within a small measurement area, and improves the accuracy and stability of weather forecasting.

Description

Technical Field

[0001] The present invention belongs to the application of global navigation satellite system technology in the field of GNSS meteorology, and particularly relates to lightweight short-term precipitation forecasting that takes into account time and parameter correlation using meteorological monitoring data and artificial intelligence models. Specifically, it relates to a lightweight short-term precipitation forecasting method and device that takes into account time and parameter correlation. Background Art

[0002] Due to the highly non-linear, random, and complex characteristics of precipitation itself, traditional numerical weather prediction methods have problems such as relying on a large amount of background data, heavy computational burden, and inaccurate prediction of small-scale weather in special regions, making it difficult to meet the application requirements of lightweight, precise, and specific small-range scenarios.

[0003] In recent years, deep learning methods can automatically and effectively extract information from a large amount of observational data, and have received extensive attention in the field of meteorology. The important application of Beidou / GNSS technology in meteorology, especially its powerful meteorological data detection ability, provides a new data source for short-term precipitation forecasting. Beidou technology can provide high-precision, high spatio-temporal resolution quasi-real-time troposphere and water vapor products, and these data are of great significance for monitoring and forecasting small and medium-scale disaster weather in complex regions.

[0004] Short-term precipitation forecasting refers to the high-resolution prediction of sudden precipitation in the short term, and has always been an important and difficult technical problem in the meteorological field. Summary of the Invention

[0005] To solve the problems that the current numerical weather prediction model has a large amount of calculation and low resolution, and cannot meet the requirements of accurate and lightweight weather forecasting for small special measurement areas, the present invention provides a lightweight short-term precipitation forecasting method and device that takes into account time and parameter correlation. By fusing Beidou meteorological monitoring data and artificial intelligence models, it can more accurately predict short-term precipitation in a small measurement area, and improve the accuracy and stability of weather forecasting.

[0006] According to one aspect of the specification of the present invention, there is provided a lightweight short-term precipitation forecasting method that takes into account time and parameter correlation, including:

[0007] Obtain precipitation and precipitation-related influencing factors of the target measurement area;

[0008] Input the precipitation and precipitation-related influencing factors into the trained time series forecasting model, and output the first precipitation forecasting data for a future preset time period;

[0009] Input the precipitation-related influencing factors into the trained time series prediction model to output the key precipitation influencing factors for a preset future period, and input the key precipitation influencing factors for the preset future period into the trained relationship model to output the second precipitation prediction data for the preset future period;

[0010] Perform weighted fitting based on the first precipitation prediction data and the second precipitation prediction data to obtain the final precipitation prediction result.

[0011] As a further technical solution, the training of the time series prediction model includes:

[0012] Construct a time series prediction model of precipitation-related influencing factors and precipitation based on the LSTM network, using the time series of precipitation and precipitation-related influencing factors in the past for a certain period as the input, and the precipitation and key precipitation influencing factors in the future for a certain period as the output;

[0013] Use the constructed training set for iterative calculation until the preset accuracy is reached, and output the trained time series prediction model.

[0014] As a further technical solution, the training of the time series prediction model also includes constructing the key precipitation influencing factors in the training set in the following manner:

[0015] Obtain the precipitation and precipitation-related influencing factor data of the measurement area, and use the partial least squares analysis method to construct a regression model between the precipitation and the precipitation-related influencing factor data;

[0016] Calculate the contribution rate of each influencing factor and sort them from high to low according to the contribution rate, and take the first N factors with the cumulative contribution rate exceeding the preset value of 85% as the key precipitation influencing factors.

[0017] As a further technical solution, the method also includes the following preprocessing of the precipitation and precipitation-related influencing factors in the measurement area:

[0018] Use the time correlation of meteorological parameters to construct a reference time-varying model of the recent daily change law, and use the data in the period before the forecast for a certain period to fit the reference time-varying model, and eliminate the outliers with the model reference value ± 3 times the mean square error as the standard;

[0019] Use the singular spectrum analysis method to interpolate the missing data, and perform normalization processing on the complete data after interpolation.

[0020] As a further technical solution, the training of the relationship model includes:

[0021] Construct a relationship model between the key precipitation influencing factors and precipitation based on the Bayesian network;

[0022] Taking the key influencing factors of precipitation observed synchronously as inputs and precipitation as the output, the optimal parameters are fitted using maximum likelihood estimation;

[0023] Output the trained relationship model.

[0024] According to one aspect of the specification of the present invention, there is provided a lightweight short-term precipitation forecasting device considering time and parameter correlation, including:

[0025] A first main module for obtaining precipitation and precipitation-related influencing factors in a target measurement area;

[0026] A second main module that inputs precipitation and precipitation-related influencing factors into a trained time series forecasting model and outputs first precipitation forecasting data for a preset future period;

[0027] A third main module for inputting the precipitation-related influencing factors into a trained time series forecasting model, outputting key precipitation influencing factors for a preset future period, inputting the key precipitation influencing factors into a trained relationship model, and outputting second precipitation forecasting data for a preset future period;

[0028] A fourth main module for performing weighted fitting based on the first precipitation forecasting data and the second precipitation forecasting data to obtain a final precipitation forecasting result.

[0029] According to one aspect of the specification of the present invention, there is provided a lightweight short-term precipitation forecasting device considering time and parameter correlation, including a memory and a processor. The memory stores program instructions executed by the processor, and the processor calls the program instructions to execute the steps of the lightweight short-term precipitation forecasting method considering time and parameter correlation.

[0030] According to one aspect of the specification of the present invention, there is provided a non-transitory computer-readable storage medium. The non-transitory computer-readable storage medium stores computer instructions, and the computer instructions cause the computer to execute the steps of the lightweight short-term precipitation forecasting method considering time and parameter correlation.

[0031] Compared with the prior art, the present invention provides a lightweight short-term precipitation forecasting method that combines Beidou meteorological monitoring and artificial intelligence considering time correlation and parameter correlation, and has the following advantages compared with conventional weather forecasting methods:

[0032] (1) The present invention makes full use of the time correlation of precipitation data and the physical relevance between precipitation and other meteorological parameters, and realizes the full utilization of effective information through the combination of the two. Therefore, more accurate and stable short-term weather forecasting can be achieved;

[0033] (2) The present invention targets single-point data modeling, utilizes Beidou meteorological monitoring data, and significantly reduces the required data compared to traditional precipitation forecasting models, thereby achieving the lightweight feature. It is particularly suitable for short-range weather forecasting in small areas based on Beidou geological disaster monitoring data.

[0034] (3) Compared with existing inventions, the present invention takes into account the correlation of parameters, fuses the BNN_MET neural network and the LSTM neural network, enabling more accurate prediction of time-series data and more accurate precipitation forecasting.

[0035] (4) The neural network fusion strategy of the present invention not only involves the direct combination of BNN_MET and LSTM, but also includes the further integration of the independent output and fused output data of the time-series forecasting model. Through this multi-level data fusion method, the present invention can make full use of the advantages of each network while compensating for the deficiencies of a single network in processing complex time-series data, thus making precipitation forecasting more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] To more clearly illustrate the technical solutions in the embodiments of the present invention or in the prior art, the following will briefly introduce the drawings used in the description of the embodiments or the prior art. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0037] Figure 1 It is a schematic flow chart of a lightweight short-term precipitation forecasting method that takes into account time and parameter correlation provided by an embodiment of the present invention.

[0038] Figure 2 It is a schematic structural diagram of a lightweight short-term precipitation forecasting device that takes into account time and parameter correlation provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0039] On the premise that Beidou / GNSS provides high-precision meteorological monitoring data, the present invention specifically proposes a lightweight short-term precipitation forecasting method that takes into account time and parameter correlation and integrates Beidou meteorological monitoring and artificial intelligence, improving the short-term precipitation forecasting accuracy in small special measurement areas and enhancing the accuracy and stability of the forecasting model.

[0040] The forecasting model provided by the present invention is a single-point forecasting model, which is particularly suitable for weather forecasting based on Beidou geological disaster monitoring data.

[0041] The content mentioned in the specification of the present invention, unless otherwise specified, can be regarded as the existing well-known technology.

[0042] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention. Additionally, the technical features in each embodiment or individual embodiment provided by the present invention can be arbitrarily combined with each other to form a new technical solution. This combination is not restricted by the order of steps and / or the structural composition mode, but must be based on the ability of those of ordinary skill in the art to implement. When the combination of technical solutions results in contradictions or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the protection scope required by the present invention.

[0043] In view of the large computational amount and low resolution of the current numerical weather prediction model, which cannot meet the requirements of precise and lightweight weather forecasting for small-scale special measurement areas, an embodiment of the present invention proposes a lightweight short-term precipitation forecasting method considering temporal correlation and parameter correlation. This method is based on Beidou meteorological monitoring and artificial intelligence, and fully utilizes the temporal correlation of precipitation data and the physical correlation between precipitation and other meteorological parameters to achieve lightweight short-term precipitation forecasting. Moreover, compared with the existing weather forecasting methods, it can more accurately represent the short-term precipitation conditions within a small measurement area.

[0044] Please refer to Figure 1 , the lightweight short-term precipitation forecasting method considering temporal and parameter correlation provided by the embodiment of the present invention includes:

[0045] Step 1, obtain precipitation and precipitation-related influencing factor data of the target measurement area.

[0046] In the embodiment of the present invention, the precipitation and precipitation-related influencing factors mainly include temperature, pressure, humidity, wind speed, wind direction, ZHD, ZWD, etc. obtained by Beidou water vapor monitoring.

[0047] Step 2, input the precipitation into the trained time series forecasting model, and output the first precipitation forecasting data for a preset future period.

[0048] In the embodiments of the present invention, the time series prediction model is a time series prediction model of precipitation and precipitation-related influencing factors constructed based on the LSTM network. Its input is the time series of precipitation and precipitation-related influencing factors for a past period of time, and the output is the precipitation and key precipitation-influencing factors for a future period of time. The relationship model is a relationship model between key precipitation-influencing factors and precipitation constructed based on the Bayesian network. During training, its input is the key precipitation-influencing factors observed synchronously (i.e., the top N factors with a cumulative contribution rate exceeding 85% mentioned above), and the output is precipitation (when applied in practice, the input is the key precipitation-influencing factors output by the LSTM).

[0049] It should be noted that the time series prediction model is constructed based on the LSTM network. It can predict the key precipitation-influencing factors for a future period based on the precipitation-related influencing factors in the past period, or predict the precipitation data for a future period based on the existing precipitation data and precipitation-related influencing factors. These two predictions can be carried out simultaneously or separately, and the present invention does not limit this.

[0050] When constructing the training set of the time series prediction model, it also includes: determining the key factors affecting precipitation according to the characteristics of the target measurement area. In the embodiments of the present invention, the partial least squares analysis method is used to construct a regression model between precipitation and multiple influencing factors, calculate the contribution rate of each influencing factor and sort them from high to low, and take the top N factors with a cumulative contribution rate exceeding 85% as the key precipitation-influencing factors, which are also the input items for subsequent training models. The factors with a 15% contribution rate are considered to have no prediction potential and are ignored as error terms.

[0051] It should be noted that the multiple influencing factors described in the embodiments of the present invention include Beidou monitoring data and meteorological monitoring data from other sources. Among them, the Beidou meteorological monitoring data includes: temperature, pressure, humidity, wind direction, wind speed, precipitation, tropospheric static delay (ZHD), and tropospheric wet delay (ZWD). The meteorological monitoring data from other sources is mainly spatio-temporal data, including: terrain, time, land cover type, and climate type classification.

[0052] Step 3, input the precipitation-related influencing factors into the trained time series prediction model, output the key precipitation-influencing factors for a future preset period, and input the key precipitation-influencing factors into the trained relationship model to output the second precipitation prediction data for a future preset period.

[0053] Specifically, taking the precipitation prediction with a three-day hourly resolution as an example based on the data with a seven-day hourly resolution in the past, the precipitation and precipitation-related influencing factors are input into the time series prediction model, and the precipitation prediction result with a three-day hourly resolution in the future is output as forecast1.

[0054] The precipitation-related influencing factors with hourly resolution in the past seven days are input into the time series forecast model to obtain the key influencing factors of precipitation with hourly resolution in the next three days. The basic meteorological parameters with hourly resolution in the next three days are then input into the relationship model to output the precipitation forecast results with hourly resolution in the next three days as forecast2.

[0055] Step 4: Perform weighted averaging based on the first precipitation forecast data forecast1 and the second precipitation forecast data forecast2 to obtain the final precipitation forecast result.

[0056] .

[0057] In the embodiment of the present invention, the construction and training process of the relationship model is as follows:

[0058] Construct a relationship model between key factors affecting precipitation and precipitation parameters. Use Bayesian neural network and historical measured meteorological data and its change rate as well as spatiotemporal data to construct a nonlinear relationship model between precipitation and key factors affecting precipitation (including Beidou meteorological monitoring data, spatiotemporal data and other key factors determined in step 1). Specifically, take the key factors affecting precipitation as input and precipitation as output, use the maximum likelihood estimation method to fit the optimal parameters, and construct a Bayesian neural network, called BNN_MET.

[0059] In the embodiment of the present invention, the construction and training process of the time series forecast model is as follows:

[0060] Construct a time series forecast model for precipitation-related influencing factors (mainly temperature, air pressure, humidity, wind speed, wind direction, ZHD, ZWD, etc. obtained from Beidou meteorological monitoring) and precipitation. Using the long-term precipitation-related influencing factor data and the time series of precipitation data as input, the LSTM model is integrated to construct a time series forecast model for precipitation-related influencing factors and precipitation. Considering the short-term forecast needs, only the hourly resolution data of the past 7 days is used to forecast the key influencing factors and precipitation of the next 3 days with hourly resolution.

[0061] Due to the existence of satellite observation errors, some data may have "outliers". Therefore, before model construction and training, the acquired data needs to be preprocessed. Specifically, in view of the "outliers" phenomenon in these data, it is necessary to model the meteorological parameters and use the time correlation of meteorological parameters to build a reference time-varying model of recent daily variation patterns. For example, the data of the previous 7 days of the forecast are used to fit the reference model, and the model reference value ±3 times the mean error is used as the standard to eliminate outliers; by analogy, the model is built with a 7-day window, and the outliers of the next day are eliminated, and the data is rolled back to the last day. Then, the singular spectrum analysis method is used to interpolate the missing data, and the complete data is normalized to prepare clean data for subsequent model construction.

[0062] The preprocessed data can be expressed as: , where represents the mean square error, represents the model reference value, represents the specific data.

[0063] When performing precipitation forecasting based on the constructed relationship model BNN_Met and the time series forecasting model LSTM_MET, the key precipitation influencing factors predicted by the LSTM_MET model are input into the BNN_Met model to obtain the precipitation forecasting results for the next 3 days, denoted as forecast2, and the precipitation predicted by the LSTM_MET model is denoted as forcast1. Then, the data for precipitation forecasting using the constructed model is compared with the measured precipitation results, and the bias Bias and standard deviation STD of forecast1 and forecast2 are statistically calculated, and then weighted average is performed according to the following formula to obtain the final forecasting result Forecast_F.

[0064] .

[0065] Finally, the above final forecasting result is evaluated and verified. When its forecasting accuracy meets the expected requirements, the above model is encapsulated and used for actual short-term and impending precipitation forecasting.

[0066] The implementation basis of each embodiment of the present invention is achieved through programmed processing by a device with processor functions. Therefore, in engineering practice, the technical solutions and their functions of each embodiment of the present invention are encapsulated into various modules. Based on this actual situation, on the basis of the above embodiments, an embodiment of the present invention provides a lightweight short-term and impending precipitation forecasting device that takes into account time and parameter correlation, and this device is used to execute the lightweight short-term and impending precipitation forecasting method that takes into account time and parameter correlation in the above method embodiments.

[0067] See Figure 2 , this device includes: a first main module for obtaining precipitation and precipitation-related influencing factors of a target measurement area; a second main module for inputting the precipitation and precipitation-related influencing factors into a trained time series forecasting model and outputting first precipitation forecasting data for a future preset period; a third main module for inputting the precipitation-related influencing factors into a trained time series forecasting model, outputting key precipitation influencing factors for a future preset period, inputting the key precipitation influencing factors into a trained relationship model, and outputting second precipitation forecasting data for a future preset period; a fourth main module for performing weighted fitting based on the first precipitation forecasting data and the second precipitation forecasting data to obtain the final precipitation forecasting result.

[0068] The lightweight short-term precipitation forecasting device considering time and parameter correlation provided by the embodiments of the present invention aims at the current situation that the current numerical weather prediction model has a large amount of calculation and low resolution, and cannot meet the needs of small-scale special measurement areas for accurate and lightweight weather forecasting. By adopting several modules in Figure 2 , through integrating Beidou meteorological monitoring data and artificial intelligence models, it can more accurately express the short-term precipitation situation within the small measurement area, improving the accuracy and stability of weather forecasting.

[0069] It should be noted that the device embodiments provided by the present invention, in addition to being used to implement the methods in the above method embodiments, are also used to implement the methods in other method embodiments provided by the present invention. The difference lies only in setting corresponding functional modules, and its principle is basically the same as that of the above device embodiments provided by the present invention. As long as those skilled in the art, based on the above device embodiments, refer to the specific technical solutions in other method embodiments, obtain corresponding technical means by combining technical features, and the technical solutions composed of these technical means, and on the premise of ensuring the practicability of the technical solutions, improve the modules in the above device embodiments to obtain corresponding device-type embodiments for implementing the methods in other method-type embodiments. For example:

[0070] Based on the content of the above device embodiments, as a preferred embodiment, the lightweight short-term precipitation forecasting device considering time and parameter correlation provided by the embodiments of the present invention further includes a relationship model construction and training module for executing the following instructions:

[0071] Construct a relationship model between the key influencing factors of precipitation and precipitation based on the Bayesian network;

[0072] Take the simultaneously observed key influencing factors as input and precipitation as output, and use maximum likelihood estimation to fit the optimal parameters;

[0073] Output the trained relationship model.

[0074] Based on the content of the above device embodiments, as a preferred embodiment, the lightweight short-term precipitation forecasting device considering time and parameter correlation provided by the embodiments of the present invention further includes a time series forecasting model construction and training module for executing the following instructions:

[0075] Construct a time series forecasting model between the precipitation-related influencing factors and precipitation based on the LSTM network, taking the time series of precipitation and precipitation-related influencing factors in the past several time periods as input and the precipitation and key influencing factors of precipitation in the future several time periods as output;

[0076] Use the constructed training set for iterative calculation until the preset accuracy is reached, and output the trained time series forecasting model.

[0077] Further, the training of the time series prediction model further includes constructing the key influencing factors of precipitation in the training set in the following manner:

[0078] Obtain the meteorological related monitoring data of the target measurement area, and use the partial least squares analysis method to construct a regression model between precipitation and multiple influencing factors;

[0079] Calculate the contribution rate of each influencing factor and sort them from high to low according to the contribution rate, and take the top N factors with the cumulative contribution rate exceeding the preset value of 85% as the key influencing factors.

[0080] Further, the training of the time series prediction model further includes the following preprocessing of the precipitation and precipitation related influencing factors in the measurement area:

[0081] Use the time correlation of meteorological parameters to construct a reference time-varying model of the recent daily variation law, and use the data of several time periods before the prediction to fit the reference time-varying model, and eliminate the outliers with the model reference value ± m times the mean square error as the standard;

[0082] Use the singular spectrum analysis method to interpolate the missing data, and perform normalization processing on the complete data after interpolation.

[0083] Based on the same inventive concept as the above embodiments, the embodiments of the present invention further provide a lightweight short-term precipitation prediction device that takes into account time and parameter correlation, including a memory and a processor. The memory stores program instructions executed by the processor, and the processor calls the program instructions to execute the steps of the lightweight short-term precipitation prediction method that takes into account time and parameter correlation.

[0084] In the embodiments of the present invention, the memory can be a non-volatile memory, such as a hard disk drive (HDD) or a solid-state drive (SSD), etc., or a volatile memory, such as a random-access memory (RAM). The memory is any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory in the embodiments of the present invention can also be a circuit or any other device capable of implementing a storage function, for storing program instructions and / or data.

[0085] In an embodiment of the present invention, the processor may be a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a field-programmable gate array, or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, and can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. The general-purpose processor may be a microprocessor or any conventional processor, etc. The steps of the method disclosed in combination with the embodiments of the present invention can be directly embodied as being executed by a hardware processor, or executed by a combination of hardware and software modules in the processor.

[0086] Based on the same inventive concept as the above embodiments, an embodiment of the present invention further provides a non-transitory computer-readable storage medium, and the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions cause the computer to execute the steps of the lightweight short-term precipitation forecasting method that takes into account time and parameter correlation.

[0087] When the above computer instructions are implemented in the form of software functional units and sold or used as independent products, they are stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention essentially, or the part that contributes to the prior art, or a part of this technical solution is embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (a personal computer, a server, or a network device) to execute all or part of the steps of the methods described in the various method embodiments of the present invention. The foregoing storage medium includes: USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical disks, and various media for storing program codes.

[0088] In summary of the above embodiments, the lightweight short-term precipitation forecasting method that combines Beidou meteorological monitoring and artificial intelligence and takes into account time and parameter correlation provided by the present invention can express the short-term precipitation situation within a small measurement area more accurately than ordinary weather forecasting models, and the precipitation situation continuously approaches the real situation, and finally improves the accuracy and stability of weather forecasting through the short-term precipitation forecasting model.

[0089] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some or all of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A lightweight short-term precipitation forecasting method considering temporal and parameter correlations, characterized in that, For short-term precipitation forecasting within a small test area, including: Obtain precipitation and precipitation-related influencing factors in the target test area; Input the precipitation and precipitation-related influencing factors into the trained time series forecasting model, and output the first precipitation forecasting data for a preset future period; Input the precipitation-related influencing factors into the trained time series forecasting model, output the key precipitation influencing factors for a preset future period, and input the key precipitation influencing factors for a preset future period into the trained relationship model to output the second precipitation forecasting data for a preset future period; the training of the relationship model includes: constructing a relationship model between the key precipitation influencing factors and precipitation based on a Bayesian network; using the simultaneously observed key precipitation influencing factors as input and precipitation as output, and fitting the optimal parameters using maximum likelihood estimation; outputting the trained relationship model; Perform weighted fitting based on the first precipitation forecasting data and the second precipitation forecasting data to obtain the final precipitation forecasting result.

2. The lightweight short-term precipitation forecasting method considering time and parameter correlation according to claim 1, characterized in that The training of the time series forecasting model includes: Construct a time series forecasting model of precipitation-related influencing factors and precipitation based on an LSTM network, using the time series of precipitation and precipitation-related influencing factors for a past period as input and the precipitation and key precipitation influencing factors for a future period as output; Perform iterative calculations using the constructed training set until the preset accuracy is reached, and output the trained time series forecasting model.

3. The lightweight short-term precipitation forecasting method considering time and parameter correlation according to claim 2, characterized in that, The training of the time series forecasting model also includes constructing the key precipitation influencing factors in the training set in the following manner: Obtain precipitation and precipitation-related influencing factors in the test area, and construct a regression model between precipitation and precipitation-related influencing factors using partial least squares analysis; Calculate the contribution rate of each influencing factor and sort them from high to low according to the contribution rate, and take the first N factors with a cumulative contribution rate exceeding the preset value of 85% as the key precipitation influencing factors.

4. The lightweight short-term precipitation forecasting method considering temporal and parameter correlations according to claim 3, characterized in that The method also includes the following preprocessing of precipitation and precipitation-related influencing factors in the test area: Construct a reference time-varying model of the recent daily variation law using the time correlation of meteorological parameters, fit the reference time-varying model using data for a period before the forecast, and eliminate outliers using the model reference value ± 3 times the mean square error as the standard; Use the singular spectrum analysis method to interpolate missing data and perform normalization processing on the complete data after interpolation.

5. A lightweight short-term precipitation forecasting device that takes into account time and parameter correlations, characterized in that, For short-term precipitation forecasting within a small test area, including: The first main module is used to obtain precipitation and precipitation-related influencing factors in the target test area; The second main module inputs the precipitation and precipitation-related influencing factors into the trained time series forecasting model and outputs the first precipitation forecasting data for a preset future period; The third main module is used to input the precipitation-related influencing factors into the trained time series forecasting model, output the key precipitation influencing factors for a preset future period, and input the key precipitation influencing factors into the trained relationship model to output the second precipitation forecasting data for a preset future period; the training of the relationship model includes: constructing a relationship model between the key precipitation influencing factors and precipitation based on a Bayesian network; using the simultaneously observed key precipitation influencing factors as input and precipitation as output, and fitting the optimal parameters using maximum likelihood estimation; outputting the trained relationship model; The fourth main module is used to perform weighted fitting based on the first precipitation forecast data and the second precipitation forecast data to obtain the final precipitation forecast result.

6. A lightweight short-term precipitation forecasting device considering time and parameter correlation, characterized in that, It includes a memory and a processor. The memory stores program instructions executed by the processor, and the processor calls the program instructions to execute the steps of the lightweight short-term precipitation forecasting method considering time and parameter correlation according to any one of claims 1 to 5.

7. A non-transitory computer-readable storage medium, characterized in that, The non-transitory computer-readable storage medium stores computer instructions, and the computer instructions cause the computer to execute the steps of the lightweight short-term precipitation forecasting method considering time and parameter correlation according to any one of claims 1 to 5.

Images