Meteorological prediction method and system based on multi-dimensional observation data fusion

Abstract

The invention provides a meteorological prediction method and system based on multi-dimensional observation data fusion, and belongs to the technical field of meteorological prediction. The method comprises the following steps: S1, calling a first-scale meteorological prediction model to output a first-scale meteorological prediction result based on first-scale observation data; s2, calling a second-scale meteorological prediction model to output a second-scale meteorological prediction result based on the second-scale observation data; s3, judging whether the first scale weather prediction result and the second scale weather prediction result have an overlapped scale result or not; if yes, adding an overlapping scale result as a forward feedback sample into a prediction sample set; otherwise, expanding the first scale observation data, increasing the first scale, and returning to the step S1; the first scale is different from the second scale. By fusing the meteorological prediction results of different scales and performing scale adjustment, the technical scheme of the invention can improve the scene adaptability and accuracy of the meteorological prediction results.

Description

Technical Field

[0001] This invention belongs to the field of meteorological forecasting technology, and particularly relates to a meteorological forecasting method and system based on multidimensional observation data fusion, a computer-readable storage medium for implementing the method, a computer program product, and an electronic device. Background Technology

[0002] Weather forecasting is based on numerical data collected by meteorological sensors or cloud images taken by satellites. It comprehensively utilizes meteorology, fluid mechanics, statistics, and various machine learning methods to make qualitative or quantitative judgments or calculate meteorological attribute values ​​for regional weather conditions in the future.

[0003] Meteorological forecasting methods are divided into extrapolation forecasting techniques and numerical model forecasting techniques. Extrapolation forecasting mainly relies on data collected by satellite radar and ground observation devices to judge the development trend of meteorological formation factors based on meteorological development mechanisms, and to make scientific predictions about rainfall status by combining expert experience. Numerical model forecasting is based on relevant historical meteorological data, and establishes a data analysis model through offline training. Combined with current meteorological data, the model calculates relevant indicators of future meteorological conditions. Examples include the multi-scale fusion lightning forecasting method based on large meteorological models and observational data proposed in Chinese invention patent application CN202410901306, and the small-scale meteorological forecasting method proposed in CN202310457853.

[0004] However, the accuracy of forecasts from different scales of weather forecasting models varies considerably. Generally speaking, traditional numerical weather prediction models are usually more effective than short-term, refined forecasts for localized areas in large-scale weather forecasts. However, in certain application areas, such as power system safety assessments and traffic safety assessments under extreme weather conditions, meteorological departments need to provide multi-scale, especially small-scale, weather forecasting services or issue weather disaster warnings with a certain degree of accuracy. Summary of the Invention

[0005] To address the aforementioned problems, this invention proposes a meteorological forecasting method and system based on multidimensional observation data fusion, a computer-readable storage medium for implementing the method, a computer program product, and an electronic device.

[0006] In a first aspect of the invention, a meteorological forecasting method based on multidimensional observation data fusion is proposed, the method comprising: S1: Based on the first-scale observation data, call the first-scale meteorological prediction model to output the first-scale meteorological prediction results; S2: Based on the second-scale observation data, call the second-scale meteorological forecast model to output the second-scale meteorological forecast results; S3: Determine whether there is any overlap between the first-scale meteorological forecast results and the second-scale meteorological forecast results; If so, the overlap scale result is added to the prediction sample set as a positive feedback sample; Otherwise, expand the observation data at the first scale and increase the first scale, then return to step S1; The first scale is different from the second scale.

[0007] Step S3 further includes: If the meteorological forecast results at the first scale and the meteorological forecast results at the second scale do not overlap, then update the observation data at the second scale, shorten the second scale, and return to step S1.

[0008] Step S3, determining whether there is any overlap between the first-scale meteorological forecast result and the second-scale meteorological forecast result, specifically includes: Obtain the intersection scale range of the first-scale meteorological forecast results and the second-scale meteorological forecast results; Determine whether there is any overlap between the first meteorological forecast result corresponding to the first scale interval and the second meteorological forecast result corresponding to the second scale interval; If so, then the meteorological forecast results at the first scale and the meteorological forecast results at the second scale have overlapping scale results; Otherwise, there will be no overlap between the first-scale meteorological forecast results and the second-scale meteorological forecast results.

[0009] When step S3 determines that there is no overlapping scale between the first-scale meteorological forecast result and the second-scale meteorological forecast result, the method further includes: Receive the latest observation data sent by meteorological observation equipment; The latest observation data is added to the first-scale observation data to obtain the expanded first-scale observation data.

[0010] When step S3 determines that there is no overlapping scale between the first-scale meteorological forecast result and the second-scale meteorological forecast result, the method further includes: Receive the latest observation data sent by meteorological observation equipment; The second-scale observation data is truncated to at least second-scale observation sub-data A and second-scale observation sub-data B; The latest observation data is fused with either the second-scale observation sub-data A or the second-scale observation sub-data B to obtain the updated second-scale observation data.

[0011] After adding the overlap scale result as a positive feedback sample to the prediction sample set in step S3, the method further includes: S4: Retrain the first-scale meteorological prediction model or the second-scale meteorological prediction model based on the prediction sample set, and return to step S2 or S1.

[0012] The first-scale or second-scale meteorological forecast results include at least one of the following meteorological parameters corresponding to different time periods: Predicting temperature and its duration, air pressure and its duration, rainfall and its duration, wind direction and its duration, wind speed and its duration, lightning intensity and its duration, snowfall intensity and its duration, etc.

[0013] In a second aspect of the present invention, in order to implement the method described in the first aspect, a meteorological forecasting system based on multidimensional observation data fusion is proposed. The meteorological forecasting system communicates with at least one meteorological observation device. The meteorological forecasting system is configured with at least a pre-trained first-scale meteorological forecasting model and a second-scale meteorological forecasting model. The system also includes: An initial prediction unit, which, based on first-scale observation data, calls a first-scale meteorological prediction model to output a first-scale meteorological prediction result, and based on second-scale observation data, calls a second-scale meteorological prediction model to output a second-scale meteorological prediction result; The result evaluation unit evaluates whether there are overlapping scale results between the first-scale meteorological forecast results and the second-scale meteorological forecast results. When the result evaluation unit assesses that there is an overlap between the first-scale meteorological forecast result and the second-scale meteorological forecast result, the sample expansion unit adds the overlapped scale result as a positive feedback sample to the forecast sample set. A model retraining unit, which retrains the first-scale meteorological prediction model or the second-scale meteorological prediction model based on the prediction sample set; The scale adjustment unit performs scale adjustment of the observation data when the result evaluation unit determines that there is no overlap between the first-scale meteorological forecast result and the second-scale meteorological forecast result.

[0014] The scaling unit performs scaling of the observed data, specifically including: Receive the latest observation data sent by meteorological observation equipment; The latest observation data is added to the first-scale observation data to obtain the expanded first-scale observation data. And / or, The second-scale observation data is truncated to at least second-scale observation sub-data A and second-scale observation sub-data B; The latest observation data is fused with either the second-scale observation sub-data A or the second-scale observation sub-data B to obtain the updated second-scale observation data.

[0015] In a third aspect of the invention, a computer-readable storage medium is also provided for storing computer instructions that, when executed on an electronic device, cause the electronic device to perform all or part of the steps of the aforementioned meteorological forecasting method based on multidimensional observation data fusion.

[0016] In a fourth aspect of the invention, a computer device is also provided, the computer device including a processor and a memory, the memory for storing instructions, and the processor for calling the instructions in the memory, causing the computer device to execute the aforementioned meteorological forecasting method based on multidimensional observation data fusion.

[0017] In a fifth aspect of the invention, a computer program product is also provided, the product comprising a computer program that, when executed, implements all or part of the steps of the aforementioned meteorological forecasting method based on multidimensional observation data fusion.

[0018] The technical solution of this invention, in the initial stage, firstly, based on the first-scale observation data, calls the first-scale meteorological prediction model to output the first-scale meteorological prediction result, and simultaneously based on the second-scale observation data, calls the second-scale meteorological prediction model to output the second-scale meteorological prediction result. Then, it is determined whether there are overlapping scale results between the first-scale meteorological prediction result and the second-scale meteorological prediction result. When there are overlapping scale results, the overlapping scale results are added as positive feedback samples to the prediction sample set for subsequent update training. Otherwise, the first-scale observation data is expanded, and the first scale is increased or the second-scale observation data is updated, and the second scale is shortened before continuing to execute the method. In this way, by fusing meteorological prediction results of different scales and performing scale adjustment, the scene adaptability and accuracy of the meteorological prediction results are improved.

[0019] Further advantages of the present invention will be further detailed in the Specific Embodiments section in conjunction with the accompanying drawings. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart of the main body of a meteorological forecasting method based on multidimensional observation data fusion according to an embodiment of the present invention; Figure 2 This is a flowchart of the main body of a meteorological forecasting method based on multidimensional observation data fusion, which is another preferred embodiment of the present invention. Figure 3 A schematic diagram illustrating the communication between a meteorological forecasting system and meteorological observation equipment that implements a meteorological forecasting method based on multidimensional observation data fusion; Figure 4 This is a schematic diagram of the functional units of a meteorological forecasting system based on multidimensional observation data fusion, according to an embodiment of the present invention. Detailed Implementation

[0023] In the specific embodiments of this application, if the embodiments of the relevant technical solutions involve user-related data, then when the embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0024] See Figure 1 , Figure 1 This is a main flowchart of a meteorological forecasting method based on multidimensional observation data fusion according to an embodiment of the present invention. The main flowchart shows three main steps S1-S3, and the specific implementation of each step is as follows: S1: Based on the first-scale observation data, call the first-scale meteorological prediction model to output the first-scale meteorological prediction results; S2: Based on the second-scale observation data, call the second-scale meteorological forecast model to output the second-scale meteorological forecast results; S3: Determine whether there is any overlap between the first-scale meteorological forecast results and the second-scale meteorological forecast results; If so, the overlapping scale results are added to the prediction sample set as positive feedback samples; at this time, it also means that both the first-scale meteorological prediction results and the second-scale meteorological prediction results contain credible prediction intervals, so prediction results of different scales can be directly published. Otherwise, expand the observation data at the first scale and increase the first scale, then return to step S1; The first scale is different from the second scale.

[0025] Preferably, the first scale is smaller than the second scale.

[0026] For example, in one scenario, the first-scale weather forecasting model could be a small-scale (short-term) weather forecasting model, which can be used to predict the weather conditions for the next 1 hour or N hours (N < 4); the second-scale weather forecasting model could be a medium-scale or large-scale (medium-to-long-term) weather forecasting model, which can be used to predict the weather conditions for the next 12 hours / the next three days / the next 7 days. In another scenario, the first-scale weather forecasting model may be, for example, a mesoscale weather forecasting model, which can be used to predict the weather conditions for the next 12 or 24 hours; the second-scale weather forecasting model may be, for example, a large-scale (long-term) weather forecasting model, which can be used to predict the weather conditions for the next three days or the next seven days.

[0027] Different application scenarios present different weather forecasting needs. For example, agricultural production may require large-scale (long-term) weather forecasts; traffic management (such as flight route scheduling) may require small-scale (short-term) weather forecasts; and other scenarios (such as power system maintenance and operation) may require mesoscale or even longer-term weather forecasts.

[0028] While traditional numerical weather prediction models perform well in forecasting large-scale weather systems, they still fall short in providing detailed short-term forecasts for localized areas. However, in specific application areas, such as power system safety assessments and traffic safety assessments under extreme weather conditions, meteorological departments are required to provide multi-scale, especially small-scale, weather forecasting services or issue weather disaster warnings with a certain level of accuracy.

[0029] Therefore, the above-mentioned technical solution of this application is proposed.

[0030] Specifically, the first-scale observation data can be the observation data obtained in the most recent 4 hours, such as the changes in the highest and lowest temperatures, wind force and direction, and the frequency of different wind force levels; the second-scale observation data can be the ground observation station, radar, satellite data, etc. obtained in the most recent 12 hours. In step S3, determining whether there is an overlap between the first-scale meteorological forecast result and the second-scale meteorological forecast result specifically includes: Obtain the intersection scale interval of the first-scale meteorological forecast result and the second-scale meteorological forecast result; determine whether there is any overlap between the first meteorological forecast result corresponding to the intersection scale interval and the second meteorological forecast result corresponding to the intersection scale interval. If so, then the meteorological forecast results at the first scale and the meteorological forecast results at the second scale have overlapping scale results; Otherwise, there will be no overlap between the first-scale meteorological forecast results and the second-scale meteorological forecast results.

[0031] For example, suppose in the first case, the first-scale weather forecast is the weather forecast for the next 4 hours (segment labeled 0-4), and the second-scale weather forecast is the weather forecast for the next 12 hours (segment labeled 0-12). Then obtain the intersection scale range of the first-scale meteorological forecast result and the second-scale meteorological forecast result, i.e., 0-4; At this point, the second-scale meteorological forecast results (the meteorological forecast results for the next 12 hours (segment marked as 0-12)) need to be divided into two time segments: (0-4) and (4-12); the intersection scale interval is (0-4). Of course, assuming the second scenario, the first-scale meteorological forecast result may also be the meteorological forecast result for the next 2 hours (segment labeled 2-4). In this case, the second-scale meteorological forecast result (the meteorological forecast result for the next 12 hours (segment labeled 0-12)) needs to be divided into three time segments: (0-2), (2-4), and (4-12). The intersection scale interval is (2-4).

[0032] Whether there is overlap between the first-scale meteorological forecast results and the second-scale meteorological forecast results specifically includes: After obtaining the intersection scale interval of the first-scale meteorological forecast result and the second-scale meteorological forecast result, it is determined whether there is any overlap between the first meteorological forecast result corresponding to the intersection scale interval and the second meteorological forecast result corresponding to the intersection scale interval. Taking the first case as an example, at this time, it is determined whether the second-scale meteorological forecast result corresponding to the intersection scale interval (0-4) overlaps with the first-scale forecast result; In practical applications, each meteorological forecast result includes the forecast results of multiple meteorological factors (predicted temperature, predicted air pressure, predicted rainfall, predicted wind direction, predicted wind speed, lightning intensity, etc.), including predicted temperature and its duration, predicted air pressure and its duration, predicted rainfall and its duration, predicted wind direction and its duration, predicted wind speed and its duration, lightning intensity and its duration, snowfall intensity and its duration, etc.

[0033] The human-computer interaction interface can display the prediction results of the above-mentioned meteorological factors in a graphical way, such as the predicted value of a certain meteorological factor and its duration, trend graph, etc. At this point, to determine whether the second-scale meteorological forecast result corresponding to the intersection scale interval (0-4) overlaps with the first-scale forecast result, it can be determined by observing whether the predicted values ​​of multiple meteorological factors and their durations and trend graphs overlap.

[0034] Preferably, if the predicted values ​​and durations of the M meteorological factors in the first-scale prediction result graph overlap with the predicted values ​​and durations of the N meteorological factors in the second-scale meteorological prediction result corresponding to the intersection scale interval (0-4), then the first-scale meteorological prediction result and the second-scale meteorological prediction result have overlapping scale results; otherwise, the first-scale meteorological prediction result and the second-scale meteorological prediction result do not have overlapping scale results.

[0035] Preferably, both M and N are integers greater than 2.

[0036] Furthermore, in Figure 1 Based on this, see Figure 2 , Figure 2 The main flowchart of a meteorological forecasting method based on multidimensional observation data fusion according to another preferred embodiment of the present invention is shown.

[0037] Figure 2 In one embodiment, step S3 further includes: If the meteorological forecast results at the first scale and the meteorological forecast results at the second scale do not overlap, then update the observation data at the second scale, shorten the second scale, and return to step S1.

[0038] Specifically, when step S3 determines that there is no overlapping scale between the first-scale meteorological forecast result and the second-scale meteorological forecast result, the method further includes: Receive the latest observation data sent by meteorological observation equipment; The second-scale observation data is truncated to at least second-scale observation sub-data A and second-scale observation sub-data B; The latest observation data is fused with either the second-scale observation sub-data A or the second-scale observation sub-data B to obtain the updated second-scale observation data.

[0039] In another aspect, when step S3 determines that there is no overlapping scale between the first-scale meteorological forecast result and the second-scale meteorological forecast result, the method further includes: Receive the latest observation data sent by meteorological observation equipment; The latest observation data is added to the first-scale observation data to obtain the expanded first-scale observation data.

[0040] After adding the overlap scale result as a positive feedback sample to the prediction sample set in step S3, the method further includes: S4: Retrain the first-scale meteorological prediction model or the second-scale meteorological prediction model based on the prediction sample set, and return to step S2 or S1.

[0041] As can be seen, the above-mentioned further preferred embodiments not only integrate meteorological forecast results at different scales and perform scale adjustment, but also further construct an enhanced forecast sample set based on credible forecast results (with overlapping results), so as to be used for subsequent model retraining, forming a self-learning closed-loop feedback process of using the model - training the model - optimizing the model - using the model again - ..., which further improves the scenario adaptability and accuracy of meteorological forecast results.

[0042] exist Figure 2-Figure 1 Based on the method implementation examples, the following will examine... Figures 3-4 .

[0043] Figure 3 A schematic diagram illustrating the communication between a meteorological forecasting system and meteorological observation equipment that implements a meteorological forecasting method based on multidimensional observation data fusion is provided.

[0044] exist Figure 3 In this system, a meteorological forecasting system based on multidimensional observation data fusion communicates with at least one meteorological observation device. The meteorological forecasting system is configured with at least a pre-trained first-scale meteorological forecasting model and a second-scale meteorological forecasting model.

[0045] Preferably, the meteorological forecasting system serves as the local forecasting platform, while the meteorological observation equipment data serves as the remote meteorological observation equipment, both located in different geographical locations. The meteorological observation equipment data can be, for example, meteorological observation aircraft, meteorological data acquisition drones, radar, anemometers, and other devices distributed across multiple different observation locations. These devices collect data according to a fixed pattern or frequency and transmit it to the local forecasting platform. Due to bandwidth limitations in data acquisition or transmission, there will always be a certain delay between the data obtained by the local forecasting platform and the meteorological observation equipment data. The local forecasting platform typically uses different meteorological models to make predictions based on the obtained observation data of different dimensions or scales. If the prediction results are unreliable (i.e., the first-scale meteorological prediction results and the second-scale meteorological prediction results do not overlap), it needs to wait for the latest data, or the local forecasting platform can directly send a command to the meteorological observation equipment to immediately send relevant observation data. However, directly sending commands to the meteorological observation equipment to immediately send relevant observation data every time also consumes data transmission resources and affects the acquisition mode and performance (e.g., battery life) of the remote meteorological observation equipment. Based on this, the technical solution of this application only performs this operation when the relevant conditions are met (i.e., there is no overlap between the first-scale meteorological forecast result and the second-scale meteorological forecast result), thus balancing work performance and result accuracy.

[0046] For this purpose, see further. Figure 4 , Figure 4 This is a schematic diagram of the functional units of a meteorological forecasting system based on multidimensional observation data fusion, according to an embodiment of the present invention. Figure 4 The system also includes: An initial prediction unit, which, based on first-scale observation data, calls a first-scale meteorological prediction model to output a first-scale meteorological prediction result, and based on second-scale observation data, calls a second-scale meteorological prediction model to output a second-scale meteorological prediction result; The result evaluation unit evaluates whether there are overlapping scale results between the first-scale meteorological forecast results and the second-scale meteorological forecast results. When the result evaluation unit assesses that there is an overlap between the first-scale meteorological forecast result and the second-scale meteorological forecast result, the sample expansion unit adds the overlapped scale result as a positive feedback sample to the forecast sample set. A model retraining unit, which retrains the first-scale meteorological prediction model or the second-scale meteorological prediction model based on the prediction sample set; The scale adjustment unit performs scale adjustment of the observation data when the result evaluation unit determines that there is no overlap between the first-scale meteorological forecast result and the second-scale meteorological forecast result.

[0047] The scaling unit performs scaling of the observed data, specifically including: Receive the latest observation data sent by meteorological observation equipment; The latest observation data is added to the first-scale observation data to obtain the expanded first-scale observation data. And / or, The second-scale observation data is truncated to at least second-scale observation sub-data A and second-scale observation sub-data B; The latest observation data is fused with either the second-scale observation sub-data A or the second-scale observation sub-data B to obtain the updated second-scale observation data.

[0048] It is understood that the relevant principles and steps of the system embodiment are consistent with those of the method embodiment, so there is no need to repeat the explanation. The two can be referenced and consulted with each other.

[0049] Other technologies, principles, algorithms, or models not elaborated in detail in this application can be found in the prior art.

[0050] The technical solution of this invention, in the initial stage, firstly, based on the first-scale observation data, calls the first-scale meteorological prediction model to output the first-scale meteorological prediction result, and simultaneously based on the second-scale observation data, calls the second-scale meteorological prediction model to output the second-scale meteorological prediction result. Then, it is determined whether there are overlapping scale results between the first-scale meteorological prediction result and the second-scale meteorological prediction result. When there are overlapping scale results, the overlapping scale results are added as positive feedback samples to the prediction sample set for subsequent update training. Otherwise, the first-scale observation data is expanded, and the first scale is increased or the second-scale observation data is updated, and the second scale is shortened before continuing to execute the method. In this way, by fusing meteorological prediction results of different scales and performing scale adjustment, the scene adaptability and accuracy of the meteorological prediction results are improved.

[0051] In the foregoing embodiments section, the present invention provides multiple embodiments, each of which can constitute an independent technical solution and may contribute to the prior art, and solve corresponding technical problems. However, it should be noted that different embodiments can be combined with each other without violating logic; at the same time, each embodiment can solve at least one technical problem, but it is not required that each individual embodiment solve multiple or all technical problems.

[0052] The foregoing has shown and described the method embodiments and systems of the present invention, but it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A meteorological forecasting method based on multidimensional observation data fusion, characterized in that, The method includes: S1: Based on the first-scale observation data, call the first-scale meteorological prediction model to output the first-scale meteorological prediction results; S2: Based on the second-scale observation data, call the second-scale meteorological forecast model to output the second-scale meteorological forecast results; S3: Determine whether there is any overlap between the first-scale meteorological forecast results and the second-scale meteorological forecast results; If so, the overlap scale result is added to the prediction sample set as a positive feedback sample; Otherwise, expand the observation data at the first scale and increase the first scale, then return to step S1; The first scale is different from the second scale.

2. The meteorological forecasting method based on multidimensional observation data fusion as described in claim 1, characterized in that, Step S3 further includes: If the meteorological forecast results at the first scale and the meteorological forecast results at the second scale do not overlap, then update the observation data at the second scale, shorten the second scale, and return to step S1.

3. The meteorological forecasting method based on multidimensional observation data fusion as described in claim 1, characterized in that, Step S3, determining whether there is any overlap between the first-scale meteorological forecast result and the second-scale meteorological forecast result, specifically includes: Obtain the intersection scale range of the first-scale meteorological forecast results and the second-scale meteorological forecast results; Determine whether there is any overlap between the first meteorological forecast result corresponding to the first scale interval and the second meteorological forecast result corresponding to the second scale interval; If so, then the meteorological forecast results at the first scale and the meteorological forecast results at the second scale have overlapping scale results; Otherwise, there will be no overlap between the first-scale meteorological forecast results and the second-scale meteorological forecast results.

4. The meteorological forecasting method based on multidimensional observation data fusion as described in claim 1, characterized in that, When step S3 determines that there is no overlapping scale between the first-scale meteorological forecast result and the second-scale meteorological forecast result, the method further includes: Receive the latest observation data sent by meteorological observation equipment; The latest observation data is added to the first-scale observation data to obtain the expanded first-scale observation data.

5. The meteorological forecasting method based on multidimensional observation data fusion as described in claim 1, characterized in that, When step S3 determines that there is no overlapping scale between the first-scale meteorological forecast result and the second-scale meteorological forecast result, the method further includes: Receive the latest observation data sent by meteorological observation equipment; The second-scale observation data is truncated to at least second-scale observation sub-data A and second-scale observation sub-data B; The latest observation data is fused with either the second-scale observation sub-data A or the second-scale observation sub-data B to obtain the updated second-scale observation data.

6. The meteorological forecasting method based on multidimensional observation data fusion as described in claim 1, characterized in that, After adding the overlap scale result as a positive feedback sample to the prediction sample set in step S3, the method further includes: S4: Retrain the first-scale meteorological prediction model or the second-scale meteorological prediction model based on the prediction sample set, and return to step S2 or S1.

7. The meteorological forecasting method based on multidimensional observation data fusion as described in claim 1, characterized in that, The first-scale or second-scale meteorological forecast results include at least one of the following meteorological parameters corresponding to different time periods: Predict temperature, air pressure, rainfall, wind direction, wind speed, and lightning intensity.

8. A meteorological forecasting system based on multidimensional observation data fusion, wherein the meteorological forecasting system communicates with at least one meteorological observation device, and the meteorological forecasting system is configured with at least a pre-trained first-scale meteorological forecasting model and a second-scale meteorological forecasting model; Its features are, The system also includes: An initial prediction unit, which, based on first-scale observation data, calls a first-scale meteorological prediction model to output a first-scale meteorological prediction result, and based on second-scale observation data, calls a second-scale meteorological prediction model to output a second-scale meteorological prediction result; The result evaluation unit evaluates whether there are overlapping scale results between the first-scale meteorological forecast results and the second-scale meteorological forecast results. When the result evaluation unit assesses that there is an overlap between the first-scale meteorological forecast result and the second-scale meteorological forecast result, the sample expansion unit adds the overlapped scale result as a positive feedback sample to the forecast sample set. A model retraining unit, which retrains the first-scale meteorological prediction model or the second-scale meteorological prediction model based on the prediction sample set; The scale adjustment unit performs scale adjustment of the observation data when the result evaluation unit determines that there is no overlap between the first-scale meteorological forecast result and the second-scale meteorological forecast result.

9. A meteorological forecasting system based on multidimensional observation data fusion as described in claim 8, characterized in that: The scaling unit performs scaling of the observed data, specifically including: Receive the latest observation data sent by meteorological observation equipment; The latest observation data is added to the first-scale observation data to obtain the expanded first-scale observation data. And / or, The second-scale observation data is truncated to at least second-scale observation sub-data A and second-scale observation sub-data B; The latest observation data is fused with either the second-scale observation sub-data A or the second-scale observation sub-data B to obtain the updated second-scale observation data.

10. A computer-readable storage medium having a computer program stored thereon, wherein, When the computer program is executed by the processor, it implements the steps of the meteorological forecasting method based on multidimensional observation data fusion as described in any one of claims 1 to 7.

Images