Intelligent grid forecasting system based on multi-source data fusion

Abstract

The invention relates to the technical field of weather forecasting, in particular to an intelligent grid forecasting system based on multi-source data fusion, which comprises a data acquisition module for extracting meteorological element factors of a target area from radar data, a satellite cloud picture and a ground observation station, and extracting topographic factors of the target area from a geographic information system; the data processing module fuses the same meteorological element factors of different sources to obtain meteorological fusion factors of the meteorological elements; the intelligent grid generation module fills the meteorological fusion factors and the topographic factors into corresponding space grids; the forecasting module inputs the meteorological fusion factor and the topographic factor of each space grid into a deep learning forecasting model to obtain initial forecasting data of the target area; and the correction module corrects the initial forecast data to obtain final forecast data. According to the method, the relevance among the multi-source data, the topographic data and the meteorological elements is considered, and the precision of intelligent grid forecasting is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of weather forecasting, and more specifically, to an intelligent grid forecasting system based on multi-source data fusion. Background Art

[0002] With the increasing and more detailed demands for weather services in all walks of life, the requirements are also getting higher and higher. Weather forecasting has changed from traditional irregular station forecasting to intelligent grid forecasting with equal longitude and latitude grids, significantly improving the forecasting accuracy and refinement level.

[0003] Existing intelligent grid forecasting technologies still have some deficiencies. For example: 1) Only focusing on single-element forecasting and ignoring the correlation between multiple elements; 2) Only using limited data sources, resulting in limited forecasting accuracy; 3) The spatio-temporal accuracy of meteorological grids is insufficient, and often only considering the meteorological element data of the region, ignoring the micro-topography data differences and unable to reflect the local characteristic differences. Therefore, there is still a large room for improvement in the forecasting accuracy of existing intelligent grid forecasting technologies. Summary of the Invention

[0004] In view of this, the present invention provides an intelligent grid forecasting system based on multi-source data fusion, which considers the correlation between multi-source data, terrain data, and various meteorological elements, and improves the accuracy of intelligent grid forecasting.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] An intelligent grid forecasting system based on multi-source data fusion, characterized by comprising:

[0007] A data acquisition module, configured to extract meteorological element factors of a target area from radar data, satellite cloud images, and observation data of ground observation stations, and extract terrain factors of the target area from a geographic information system;

[0008] A data processing module, configured to fuse the same meteorological element factors from different sources respectively to obtain meteorological fusion factors for each meteorological element;

[0009] An intelligent grid generation module, configured to perform spatial grid division on the target area, fill each meteorological fusion factor and terrain factor into the corresponding spatial grid, and the meteorological fusion factors of each spatial grid change dynamically with time;

[0010] A forecasting module, configured to input the meteorological fusion factors and terrain factors of each spatial grid into a pre-trained deep learning forecasting model to predict multiple meteorological element data of the target area in a future time period, and obtain initial forecasting data;

[0011] A correction module for correcting the initial forecast data according to the correlation between different meteorological elements to obtain the final forecast data.

[0012] Furthermore, the meteorological element data includes any several or all of temperature, air pressure, wind speed, humidity, rain and snow, visibility, haze, and hail; the terrain factor is the average altitude of the real geographical location corresponding to each spatial grid.

[0013] Furthermore, the data processing module includes:

[0014] A weight determination unit for assigning different weights to the same meteorological element factors from three sources according to the prediction accuracy differences of different meteorological element factors by radar data, satellite cloud images, and ground observation stations;

[0015] A fusion unit for performing time calibration on the data from the three sources and then performing weighted summation on the same meteorological element factors from the three sources at the same time period under the same spatial grid to obtain the meteorological fusion factors of each meteorological element.

[0016] Furthermore, the calculation formula for the meteorological fusion factor is:

[0017]

[0018] where represents the meteorological fusion factor of the i-th meteorological element at the j-th time period in a certain spatial grid; represents the i-th meteorological element factor extracted from the satellite cloud image at the j-th time period in a certain spatial grid;

[0019] represents the i-th meteorological element factor extracted from the radar data at the j-th time period in a certain spatial grid; represents the i-th meteorological element factor extracted from the observed data of the ground observation station at the j-th time period in a certain spatial grid; α 1 、α 2 and α 3 respectively represent weights, and for different meteorological element factors, α 1 、α 2 and α 3 have different values.

[0020] Furthermore, the intelligent grid generation module includes:

[0021] A grid division unit for dividing the target area into uniform or non-uniform grids according to the forecast requirements as a map layer, and copying the same number of map layers for partition display according to the number of meteorological fusion factors, with each meteorological fusion factor corresponding to a partition map layer;

[0022] Filling unit, used to fill the values of each meteorological fusion factor onto the map layer of the corresponding partition, and label the corresponding terrain on each spatial grid;

[0023] Meteorological curve drawing unit, used to draw the change curves of each meteorological fusion factor on different spatial grids of the map layer during the current period;

[0024] Dynamic display unit, used to dynamically display the change of the corresponding meteorological fusion factor curve with time on the map layer of each partition.

[0025] Further, the grid division unit first divides the priorities of the attention levels of different sub-regions in the target area, and then determines the grid scale of each sub-region according to the priority order of attention levels.

[0026] Further, the forecasting module includes:

[0027] Preprocessing unit, used to compare the terrain data differences between any spatial grid and its eight neighboring spatial grids to determine the terrain morphology of this spatial grid;

[0028] Deep learning module, used to comprehensively analyze the meteorological fusion factors and terrain morphology of each spatial grid by using a pre-trained deep learning forecasting model to obtain the forecasting data of different spatial grids in the target area for the future period.

[0029] Further, the way for the preprocessing unit to determine the terrain morphology of any spatial grid is:

[0030] For land areas, if the altitude differences between a certain spatial grid and its eight neighboring spatial grids are all less than the preset value, the terrain morphology of this spatial grid is regarded as flat land;

[0031] If the altitude of a certain spatial grid is higher than that of its eight neighboring spatial grids, and the exceeded height is greater than the preset value, the terrain morphology of this spatial grid is regarded as a hill;

[0032] If the altitude of a certain spatial grid is lower than that of its eight neighboring spatial grids, and the lower height is greater than the preset value, the terrain morphology of this spatial grid is regarded as a valley.

[0033] Further, the ways for the correction module to correct each meteorological element in the initial forecasting data include:

[0034] Collect a large amount of historical meteorological data. For any meteorological element, determine other meteorological elements associated with it. Taking this meteorological element as the dependent variable and other meteorological elements as the independent variables, establish a multiple linear or nonlinear regression model;

[0035] Substitute the initial forecast values of other meteorological elements associated with this meteorological element in the initial forecast data into a multiple linear or non-linear regression model to obtain the intermediate forecast value of this meteorological element;

[0036] Take the average of the initial forecast value and the intermediate forecast value of this meteorological element in the initial forecast data as the final forecast value of this meteorological element.

[0037] Further, for the rain and snow data in the initial forecast data, the way the correction module corrects it also includes:

[0038] Evenly divide the future time period into n sub-time periods. The correction method for the precipitation / snowfall in any sub-time period i is:

[0039] P ts =P t ×P′ ts / (P′ t1 +P′ t2 +…+P′ tn )

[0040] where P ts represents the corrected precipitation / snowfall in the future sub-time period t s ; P t represents the total precipitation / snowfall in the future time period t predicted by the deep learning forecast model; P′ ts represents the precipitation / snowfall before correction in the future sub-time period t s , s = 1, 2, …, n.

[0041] As can be seen from the above technical solutions, compared with the prior art, the present invention has the following beneficial effects:

[0042] 1. By considering the meteorological element data from three sources, namely radar data, satellite cloud images, and ground observation stations, and combining the forecast accuracies of different data sources for different meteorological elements, the present invention fuses the meteorological elements from the three sources, solving the technical problem of limited forecast accuracy of a single data source.

[0043] 2. By making a refined grid division of the target area and dynamically displaying the variation of each meteorological fusion factor over time on the corresponding spatial grid, the present invention realizes the intuitive display and dynamic tracking of meteorological elements.

[0044] 3. Considering the influence of the topographic morphology differences of different spatial grids on the weather, the present invention can better reflect the regional weather characteristics and realizes the accurate forecast of local climate.

[0045] 4. Considering the correlation between different meteorological elements, the present invention corrects the initial forecast data, further improving the accuracy of weather forecasting. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to the provided drawings.

[0047] Figure 1 It is a structural block diagram of an intelligent grid forecasting system based on multi-source data fusion provided by the present invention;

[0048] Figure 2 It is a structural block diagram of the intelligent grid generation module provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0049] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0050] As Figure 1 shown, the embodiments of the present invention disclose an intelligent grid forecasting system based on multi-source data fusion, including:

[0051] A data acquisition module for extracting meteorological element factors of the target area from radar data, satellite cloud images and observation data of ground observation stations, and extracting terrain factors of the target area from a geographic information system;

[0052] A data processing module for fusing the same kind of meteorological element factors from different sources to obtain meteorological fusion factors for each meteorological element;

[0053] An intelligent grid generation module for dividing the target area into spatial grids, filling each meteorological fusion factor and terrain factor into the corresponding spatial grids, and the meteorological fusion factors of each spatial grid change dynamically with time;

[0054] A forecasting module for inputting the meteorological fusion factors and terrain factors of each spatial grid into a pre-trained deep learning forecasting model to predict multiple meteorological element data for the future period of the target area to obtain initial forecasting data;

[0055] A correction module for correcting the initial forecasting data according to the correlation between different meteorological elements to obtain the final forecasting data.

[0056] The following further explains each of the above modules.

[0057] (1) The data acquisition module extracts meteorological element factors of the target area from radar data, satellite cloud images, and observation data of ground observation stations.

[0058] Observation values of conventional meteorological elements such as air temperature, humidity, wind speed, wind direction, and air pressure, as well as information on special meteorological elements such as precipitation, visibility, and cloud amount, are extracted from the observation data of ground meteorological stations. These data have the characteristics of high spatio-temporal resolution and can reflect the actual state of the near-surface atmosphere, providing accurate boundary conditions and initial field information for forecasting.

[0059] Image processing technology is used to analyze satellite cloud images to extract characteristic information such as the distribution, shape, texture, and brightness of clouds, as well as physical parameters such as cloud top temperature and water vapor content. By dynamically monitoring and analyzing cloud images, the movement and development of weather can be tracked, such as the path of a typhoon and the position of a front, providing macroscopic weather situation information for forecasting.

[0060] Characteristic information such as echo intensity, radial velocity, and spectral width, as well as microphysical characteristics such as the phase state, size, and distribution of precipitation particles, are extracted from radar data. Radar data has high spatio-temporal resolution and detection accuracy, and can real-time monitor the internal structure and evolution process of precipitation systems, providing an important basis for short-term and nowcasting.

[0061] The finally obtained meteorological element data includes any several or all of air temperature, air pressure, wind speed, humidity, rain and snow, visibility, haze, and hail.

[0062] The data acquisition module also extracts topographic factors of the target area from the geographic information system. The topographic factors are the average altitude of the true geographical locations corresponding to each spatial grid. For example, topographic factors such as altitude, slope, and aspect will have a significant impact on meteorological elements such as air temperature and precipitation.

[0063] (2) The data processing module fuses the same type of meteorological element factors from different sources to obtain meteorological fusion factors for each meteorological element, specifically including:

[0064] The weight determination unit is used to assign different weights to the same type of meteorological element factors from three sources according to the accuracy differences in predicting different meteorological element factors by radar data, satellite cloud images, and ground observation stations;

[0065] The prediction accuracies of the three data sources for different meteorological elements have their own advantages. For example, for precipitation, the detection data of ground observation stations have the largest weight, the radar observation data have the second largest weight, and the satellite cloud image detection data have the smallest weight. For typhoon prediction, the satellite cloud image detection data have the largest weight, the radar observation data have the second largest weight, and the detection data of ground observation stations have the smallest weight. For short-term and imminent rainfall, the radar observation data have the largest weight, the satellite cloud image detection data have the second largest weight, and the data of ground observation stations have the smallest weight.

[0066] The fusion unit is used to perform time calibration on the data from the three sources, and then perform weighted summation on the same type of meteorological element factors from the three sources in the same spatial grid at the same time period to obtain the meteorological fusion factor of each meteorological element. The calculation formula for the meteorological fusion factor is:

[0067]

[0068] Where, represents the meteorological fusion factor of the i-th meteorological element in the j-th time period in a certain spatial grid; represents the i-th meteorological element factor extracted from the satellite cloud image in the j-th time period in a certain spatial grid; represents the i-th meteorological element factor extracted from the radar data in the j-th time period in a certain spatial grid; represents the i-th meteorological element factor extracted from the observation data of the ground observation station in the j-th time period in a certain spatial grid; α 1 、α 2 and α 3 respectively represent weights. For different meteorological element factors, the values of α 1 、α 2 and α 3 are different. If for a certain spatial grid, only the detection data of the satellite cloud image and the radar exist in this spatial grid, then the observation data of the ground observation station are taken as 0.

[0069] By performing weighted summation on the meteorological elements from the three sources, more reliable decision-making results can be obtained.

[0070] (3) The intelligent grid generation module divides the target area into spatial grids, fills each meteorological fusion factor and terrain factor into the corresponding spatial grids, and the meteorological fusion factors of each spatial grid change dynamically with time. As Figure 2 shown, the intelligent grid generation module specifically includes:

[0071] The grid division unit is used to divide the target area into uniform or non-uniform grids according to the forecast requirements, serving as map layers, and replicating the same number of map layers for zonal display according to the number of meteorological fusion factors, with each meteorological fusion factor corresponding to a map layer in a partition. Specifically, the grid division unit first divides the priorities of the attention levels of different sub-areas in the target area, and then determines the grid scales of each sub-area according to the priority order of the attention levels. For example, a finer grid division can be adopted for the key sub-areas of concern, while a coarser grid can be used for the relatively less important sub-areas. The size of the grid usually ranges from several kilometers to even dozens of meters.

[0072] The filling unit is used to fill the values of each meteorological fusion factor onto the map layers corresponding to the partitions, and mark the corresponding terrain labels on each spatial grid.

[0073] The meteorological curve drawing unit is used to draw the change curves of each meteorological fusion factor on different spatial grids of the map layer at the current time period.

[0074] The dynamic display unit is used to dynamically display the change of the corresponding meteorological fusion factor curve over time on the map layers of each partition.

[0075] The present invention can dynamically display the dynamic changes of meteorological factors in each spatial grid on the map layer. By zonal display of the changes of each meteorological factor and marking the terrain labels on each spatial grid, the data presentation is more intuitive.

[0076] (4) The forecasting module inputs the meteorological fusion factors and terrain factors of each spatial grid into a pre-trained deep learning forecasting model to predict the data of multiple meteorological elements in the future time period of the target area, and obtains the initial forecast data. Specifically, it includes:

[0077] The preprocessing unit is used to compare the terrain data differences between any spatial grid and its eight neighboring spatial grids to determine the terrain morphology of the spatial grid. The method for the preprocessing unit to determine the terrain morphology of any spatial grid is as follows:

[0078] For land areas, if the altitude differences between a certain spatial grid and its eight neighboring spatial grids are all less than the preset value, the terrain morphology of this spatial grid is regarded as flat land;

[0079] If the altitude of a certain spatial grid is higher than that of its eight neighboring spatial grids, and the exceeded height is greater than the preset value, the terrain morphology of this spatial grid is regarded as a hill;

[0080] If the altitude of a certain spatial grid is lower than that of its eight neighboring spatial grids, and the lower height is greater than the preset value, the terrain morphology of this spatial grid is regarded as a valley.

[0081] A deep learning module for comprehensively analyzing meteorological fusion factors and terrain morphologies of each spatial grid using a pre-trained deep learning forecasting model to obtain forecasting data of different spatial grids in the target area for a future period.

[0082] When training the deep learning model, each training sample in the training sample set includes micro-topographic data, meteorological element data of an observation area, and the true weather forecast for a future period corresponding to the meteorological element data. The observation area is grid-shaped, and its scale ranges from dozens of meters to several kilometers.

[0083] In addition, the deep learning forecasting model can also consider geographical locations. When training, category labels are assigned to each observation area, such as cities, mountainous areas, suburbs, forests, etc. For example, in urban areas, due to the influence of the heat island effect, the temperature may be higher than that of the surrounding suburbs. In mountainous areas, due to the uplifting effect of the terrain, the precipitation may be greater than that in plain areas.

[0084] (5) The correction module corrects the initial forecasting data according to the correlation between different meteorological elements to obtain the final forecasting data. The specific correction methods include:

[0085] 1) Collect a large amount of historical meteorological data. For any meteorological element, determine other meteorological elements associated with it. Taking this meteorological element as the dependent variable and other meteorological elements as independent variables, establish a multiple linear or non-linear regression model.

[0086] For example, there are correlations between temperature and air pressure, wind speed, humidity, rain and snow, visibility, haze, etc. Usually, at the same altitude, when the temperature is high, the air pressure is low, and vice versa. When heated, the movement of air molecules speeds up, the distance between molecules increases, resulting in a decrease in air pressure; when cooled, the molecular movement slows down, the distance between molecules shrinks, and the air pressure increases.

[0087] The change in temperature will affect the wind speed. For example, during the day, the ground heats up quickly under solar radiation, the temperature of the near-surface atmosphere rises, the air pressure decreases, and it is easy to form an updraft, and the wind speed may increase accordingly; at night, the ground cools down quickly, the temperature of the near-surface atmosphere decreases, the air pressure increases, the airflow is relatively stable, and the wind speed may decrease.

[0088] When the temperature rises and the water vapor content in the air remains unchanged, the saturated water vapor pressure will increase and the relative humidity will decrease; conversely, when the temperature decreases, the relative humidity will increase.

[0089] Temperature is one of the important factors affecting the formation of rain and snow. When the temperature is below 0°C, the precipitation form is usually solid precipitation such as snow or sleet; when the temperature is above 0°C, the precipitation is mostly rain. At the same time, the change in temperature will also affect the melting speed of rain and snow, etc.

[0090] An increase in temperature may enhance air convection, accelerate the diffusion of pollutants, and potentially improve visibility. However, under some special circumstances, such as the presence of an inversion layer, an increase in temperature may lead to an intensification of the inversion layer, making it difficult for pollutants to disperse and causing visibility to decrease instead. Similarly, temperature has an important impact on the formation and dissipation of haze. Low-temperature and stable weather conditions are not conducive to the diffusion of pollutants and are prone to the formation of haze. When the temperature rises and air convection strengthens, it is conducive to the dissipation of haze.

[0091] For another example, visibility is strongly correlated with factors such as haze, rain, snow, and hail. Pollutant particles, water droplets, etc. in the haze will scatter and absorb light, making visibility poor. When the haze is severe, visibility may be only dozens of meters or even lower.

[0092] The pressure changes, temperature changes, etc. before and after rain or snow weather will also indirectly affect visibility. For example, changes such as a decrease in air pressure and an increase in air humidity before rain or snow may lead to a gradual decrease in visibility.

[0093] Although the impact of hail itself on visibility is relatively small, during the formation process of hail clouds and when hail falls, it is often accompanied by other weather phenomena such as rainfall and cloud fog, which will indirectly affect visibility.

[0094] For another example, humidity is strongly correlated with meteorological elements such as rain, snow, visibility, and haze. The greater the relative humidity, the higher the water vapor content in the air, and the more conducive it is to the condensation of water vapor into water droplets or ice crystals, thus forming precipitation weather such as rain and snow.

[0095] Higher humidity will make the air humid and prone to the formation of fog, thus reducing visibility; while lower humidity is conducive to maintaining better visibility.

[0096] Humidity has a certain impact on the formation and persistence of haze. In a high-humidity environment, pollutants are prone to absorb moisture and grow, and are not easy to disperse, which will aggravate the degree and duration of haze; while in a dry environment, haze is relatively easy to dissipate.

[0097] 2) After the multiple linear or non-linear regression model, substitute the initial forecast values of other meteorological elements associated with this meteorological element in the initial forecast data into the multiple linear or non-linear regression model to obtain the intermediate forecast value of this meteorological element;

[0098] 3) Take the average of the initial forecast value and the intermediate forecast value of this meteorological element in the initial forecast data as the final forecast value of this meteorological element.

[0099] In addition, for the rain and snow data in the initial forecast data, the way the correction module corrects it can also be:

[0100] Evenly divide the future time period into n sub-time periods. The correction method for the precipitation / snowfall amount in any sub-time period i is:

[0101] P ts = P t × P′ ts / (P′ t1 + P′ t2 + … + P′ tn )

[0102] Wherein, P ts represents the corrected precipitation / snowfall amount in the future sub-time period t s ; P t represents the total precipitation / snowfall amount in the future time period t predicted by the deep learning prediction model; P′ ts represents the precipitation / snowfall amount before correction in the future sub-time period t s , s = 1, 2, …, n.

[0103] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the various embodiments can be referred to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple. For the relevant parts, refer to the description in the method section.

[0104] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An intelligent grid forecasting system based on multi-source data fusion, characterized in that: include: A data acquisition module is used to extract meteorological factors of the target area from radar data, satellite cloud images and observation data of ground observation stations, and to extract terrain factors of the target area from the geographic information system; The data processing module is used to fuse the same meteorological element factors from different sources to obtain the meteorological fusion factors of each meteorological element; The intelligent grid generation module is used to divide the target area into spatial grids, fill each meteorological fusion factor and terrain factor into the corresponding spatial grid, and the meteorological fusion factor of each spatial grid changes dynamically over time; The forecast module is used to input the meteorological fusion factors and terrain factors of each spatial grid into the pre-trained deep learning forecast model, predict multiple meteorological element data in the future period of the target area, and obtain initial forecast data; The correction module is used to correct the initial forecast data according to the correlation between different meteorological elements to obtain the final forecast data.

2. The intelligent grid forecasting system based on multi-source data fusion according to claim 1 is characterized in that: The meteorological element data include: any or all of temperature, air pressure, wind speed, humidity, rain, snow, visibility, haze and hail; the terrain factor is the average altitude of the real geographical location corresponding to each spatial grid.

3. The intelligent grid forecasting system based on multi-source data fusion according to claim 1 is characterized in that: The data processing module comprises: A weight determination unit is used to assign different weights to the same meteorological element factors from three sources according to the differences in the accuracy of the predictions of different meteorological element factors by radar data, satellite cloud images and ground observation stations; The fusion unit is used to perform time calibration on the data from the three sources, and then perform weighted summation of the same meteorological element factors from the three sources in the same time period under the same spatial grid to obtain the meteorological fusion factor of each meteorological element.

4. The intelligent grid forecasting system based on multi-source data fusion according to claim 3 is characterized in that: The calculation formula of the meteorological fusion factor is: in, It represents the meteorological fusion factor of the i-th meteorological element in a certain spatial grid in the j-period; It represents the i-th meteorological element factor extracted from the satellite cloud image at a certain spatial grid in the j-period; It represents the i-th meteorological element factor extracted from radar data in a certain spatial grid at time period j; It represents the i-th meteorological factor extracted from the observed data of the ground observation station in a certain spatial grid in the j-period; α1, α2 and α3 represent weights respectively. For different meteorological factors, the values ​​of α1, α2 and α3 are different.

5. The intelligent grid forecasting system based on multi-source data fusion according to claim 1 is characterized in that: The intelligent grid generation module comprises: The grid division unit is used to divide the target area into uniform or non-uniform grids as map layers according to the forecast requirements, and copy the same number of map layers for partition display according to the number of meteorological fusion factors. Each meteorological fusion factor corresponds to a partition map layer. A filling unit is used to fill the values ​​of each meteorological fusion factor into the map layer of the corresponding partition, and mark the corresponding terrain label on each spatial grid; The meteorological curve drawing unit is used to draw the change curve of each meteorological fusion factor on different spatial grids on the map layer in the current period; The dynamic display unit is used to dynamically display the changes of the corresponding meteorological fusion factor curve over time on the map layer of each partition.

6. The intelligent grid forecasting system based on multi-source data fusion according to claim 5 is characterized in that: The grid division unit first prioritizes the attention levels of different sub-regions in the target region, and then determines the grid scale of each sub-region according to the priority order of the attention levels.

7. The intelligent grid forecasting system based on multi-source data fusion according to claim 1 is characterized in that: The forecast module comprises: A preprocessing unit, used for comparing the terrain data difference between any spatial grid and the spatial grids of its eight neighbors, and determining the terrain morphology of the spatial grid; The deep learning module is used to use a pre-trained deep learning forecasting model to conduct a comprehensive analysis of the meteorological fusion factors and terrain morphology of each spatial grid, and obtain forecast data for different spatial grids in the target area in future periods.

8. The intelligent grid forecasting system based on multi-source data fusion according to claim 7 is characterized in that: The pre-processing unit determines the terrain morphology of any spatial grid in the following manner: For land areas, if the altitude differences between a spatial grid and its eight neighboring spatial grids are less than the preset value, the terrain of the spatial grid is considered to be flat land; If the altitude of a certain spatial grid is higher than the altitude of its eight neighboring spatial grids, and the excess altitude is greater than a preset value, the terrain of the spatial grid is considered to be hilly; If the altitude of a spatial grid is lower than the altitude of its eight neighboring spatial grids, and the lower altitude is greater than the preset value, the terrain morphology of the spatial grid is regarded as a valley.

9. The intelligent grid forecasting system based on multi-source data fusion according to claim 1 is characterized in that: The correction module corrects each meteorological element in the initial forecast data in the following manner: Collect a large amount of historical meteorological data, determine other meteorological elements associated with any meteorological element, take the meteorological element as the dependent variable and other meteorological elements as independent variables, and establish a multivariate linear or nonlinear regression model; Substituting the initial forecast values ​​of other meteorological elements associated with the meteorological element in the initial forecast data into a multivariate linear or nonlinear regression model to obtain an intermediate forecast value of the meteorological element; The initial forecast value and the intermediate forecast value of the meteorological element in the initial forecast data are averaged as the final forecast value of the meteorological element.

10. The intelligent grid forecasting system based on multi-source data fusion according to claim 1 is characterized in that: The correction module further corrects the rain and snow data in the initial forecast data in the following manner: Divide the future period t into n sub-periods uniformly. For any sub-period t s The correction method for precipitation / snowfall is: P ts =P t ×P′ ts / (P′ t1 +P′ t2 +…+P′ tn ) Among them, P ts represents the future sub-period t s Corrected precipitation / snowfall; P t represents the total precipitation / snowfall in the future period t predicted by the deep learning forecasting model; P′ ts represents the future sub-period t s Precipitation / snowfall before correction, s=1,2,…,n.

Images