A Meteorological Decision Support and Optimization Method and System Based on Artificial Intelligence

Abstract

This invention relates to an artificial intelligence-based meteorological decision support and optimization method and system. The AI-based meteorological decision support includes, upon receiving a meteorological decision task, determining the corresponding location information and task characteristics. Based on the location information and task characteristics, a personalized target data range for the current meteorological decision task is determined for the AI ​​model. Only meteorological data within the target data range is input into the corresponding AI model to obtain meteorological decision support data. In some implementations, this can improve the effectiveness of the data input to the AI ​​model, helping to reduce resource consumption.

Description

Technical Field

[0001] This invention relates to the field of meteorological data, and in particular to a meteorological decision support and optimization method and system based on artificial intelligence. Background Technology

[0002] Meteorological data not only provides guidance for the general public's daily life and travel, but also offers data support in areas such as administration and management. For example, meteorological data can provide references for disaster prevention and assessment, enabling relevant departments to issue timely warnings and make informed decisions. Compared to weather forecasts for ordinary users, meteorological decision support requires a much larger volume of meteorological data. Meteorological decision support often relies on artificial intelligence models to analyze meteorological data. Large amounts of meteorological data are input into these models to obtain meteorological decision support data that can provide reference for meteorological decisions. Current artificial intelligence models often acquire global meteorological data or input data from a fixed range, requiring significant computing power in a short period of time. This demand for computing power is also high when there are many meteorological decision-making tasks. Summary of the Invention

[0003] The purpose of this invention is to propose a meteorological decision support and optimization method and system based on artificial intelligence, in order to solve the problem that current artificial intelligence models often need to acquire global meteorological data or input a large amount of meteorological data within a fixed range when performing meteorological decision-making tasks, which easily leads to the input of a lot of invalid data and consumes a lot of computing power in a short period of time.

[0004] The artificial intelligence-based meteorological decision support and optimization method of the present invention includes:

[0005] Upon receiving a meteorological decision-making task, the location information and task characteristics corresponding to the meteorological decision-making task are determined;

[0006] Based on the location information and task characteristics, the target data range for the current meteorological decision-making task of the artificial intelligence model is determined in a personalized manner.

[0007] Meteorological data within the target data range is input into the corresponding artificial intelligence model to obtain meteorological decision support data.

[0008] Optionally, the method of determining the target data range of the artificial intelligence model for the current meteorological decision-making task in a personalized manner based on the location information and the task characteristics includes at least one of the following:

[0009] Based on the location information and task characteristics of the meteorological decision-making task, a candidate data range matching the meteorological decision-making task is retrieved from the pre-stored data range information and used as the target data range;

[0010] The target data range of the meteorological decision-making task is dynamically calculated based on the meteorological characteristics of the location corresponding to the meteorological decision-making task and / or the task characteristics.

[0011] Optionally, before individually determining the target data range for the current meteorological decision-making task using the artificial intelligence model, the method further includes:

[0012] During periods of idle computing power, multiple event simulations are performed based on the historical event data to obtain simulation data. Each event simulation inputs historical meteorological data with different data ranges into the artificial intelligence model.

[0013] The inferred data is compared with the historical event data;

[0014] The range of candidate data is determined based on the comparison results between the inferred data and the historical event data;

[0015] The candidate data range is associated with and stored in relation to the location information and features corresponding to historical event data.

[0016] Optionally, the step of obtaining the extrapolated data by performing multiple event extrapolations based on the historical event data includes: starting from the basic data range and continuously expanding the data range in the event extrapolation;

[0017] The step of determining the candidate data range based on the comparison results between the inferred data and the historical event data includes: determining the optimal data range as the candidate data range based on the trend of similarity changes between each inferred data and the historical event data.

[0018] Optionally, determining the optimal data range as the candidate data range based on the similarity trend between each simulation data and the historical event data includes:

[0019] Several data ranges are identified where the similarity between the inferred data and the historical event data is greater than the minimum similarity threshold, and these data ranges are sorted from low to high according to the amount of data.

[0020] After each increase in data volume, it is determined whether the increase in similarity between the inferred data and the historical event data is less than the minimum increase threshold. When it is determined that the increase in similarity between the (N+1)th inferred data and the historical event data compared to the Nth time is less than the minimum increase threshold, the data range used in the Nth inferred data is taken as the candidate data range.

[0021] Optionally, dynamically calculating the target data range for the meteorological decision-making task based on the meteorological characteristics and / or task features of the location corresponding to the meteorological decision-making task includes:

[0022] Obtain the basic location range for the meteorological decision-making task;

[0023] Determine the influencing factors of the location range and the corresponding influence coefficients of each influencing factor;

[0024] Multiply the basic location range by each of the influence coefficients to obtain the location range within the target data range.

[0025] Optionally, dynamically calculating the target data range for the meteorological decision-making task based on the meteorological characteristics and / or task features of the location corresponding to the meteorological decision-making task includes:

[0026] Obtain the basic time range for the meteorological decision-making task;

[0027] Determine the rate of change of the meteorological elements corresponding to the meteorological decision-making task;

[0028] Based on the aforementioned basic time range, a time range within the target data range is determined based on the rate of change of the meteorological elements, wherein the length of the time range is negatively correlated with the rate of change of the meteorological elements.

[0029] Optionally, the step of individually determining the target data range for the current meteorological decision-making task by the artificial intelligence model based on the location information and the task characteristics includes:

[0030] Based on the location information and task characteristics of the meteorological decision-making task, query from the pre-stored data range information whether there is a candidate data range that matches the meteorological decision-making task;

[0031] After finding a matching range of candidate data stored, the matching range of candidate data is determined as the target data range.

[0032] If no matching alternative data range is found, the target data range of the meteorological decision-making task is dynamically calculated based on the meteorological characteristics of the location corresponding to the meteorological decision-making task and / or the task characteristics.

[0033] Optionally, the target data range includes a location range and a time range described in relative terms;

[0034] The location range indicates the geographic radius relative to the location information;

[0035] The time range indicates the time span relative to the time corresponding to the meteorological decision-making task.

[0036] On the other hand, the present invention also provides a meteorological decision support and optimization system based on artificial intelligence, comprising:

[0037] The task parsing unit is used to determine the location information and task characteristics corresponding to the meteorological decision-making task after receiving the meteorological decision-making task.

[0038] The data range determination unit is used to determine the target data range of the artificial intelligence model for the current meteorological decision-making task in a personalized manner based on the location information and the task characteristics.

[0039] The data transmission unit is used to input only the meteorological data within the target data range into the corresponding artificial intelligence model;

[0040] Several artificial intelligence models are used to determine meteorological decision support data based on the input meteorological data.

[0041] The artificial intelligence-based meteorological decision support and optimization method provided by this invention determines the target data range individually for meteorological decision-making tasks, avoiding the need to import full data or a fixed-size dataset into the AI ​​model for computation each time. Full data or a fixed-size dataset typically contains a large amount of data that is not significantly helpful for meteorological decision-making. Because the target data range is individually determined based on the actual situation of the meteorological decision-making task, in some implementation processes, the effectiveness of the data input to the AI ​​model can be improved, helping to reduce resource consumption. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of the basic process of the meteorological decision support and optimization method according to Embodiment 1 of the present invention;

[0043] Figure 2 This is a flowchart illustrating the calculation of the candidate data range in Embodiment 2 of the present invention;

[0044] Figure 3 This is a schematic diagram of the event deduction and comparison process in Embodiment 2 of the present invention;

[0045] Figure 4 This is a flowchart illustrating the process of determining the target data range in Embodiment 3 of the present invention;

[0046] Figure 5 This is a flowchart illustrating the process of determining the target data range in Embodiment 4 of the present invention;

[0047] Figure 6 This is a schematic diagram of the structure of the artificial intelligence-based meteorological decision support and optimization system according to Embodiment 5 of the present invention. Detailed Implementation

[0048] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0049] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus. Example

[0050] Current meteorological decision support processes typically input all data or a fixed range of large amounts of data into relevant artificial intelligence models. This requires significant computing power in a short period, and the demand is even greater when there are many meteorological decision-making tasks. Therefore, this embodiment provides an AI-based meteorological decision support and optimization method that can individually determine the data range input into the AI ​​model according to the actual needs of the meteorological decision-making task. This avoids inputting all data or a fixed-scale large amount of data into the AI ​​model every time, and helps reduce the consumption of computing resources in a short period of time during some implementations.

[0051] See Figure 1 The artificial intelligence-based meteorological decision support and optimization method in this embodiment includes, but is not limited to, the following steps:

[0052] S101. After receiving the meteorological decision-making task, determine the location information and task characteristics corresponding to the meteorological decision-making task;

[0053] The location information of the meteorological decision-making task reflects the geographical location corresponding to the current meteorological decision-making task. Task characteristics include other information besides geographical location in the meteorological decision-making task, such as, but not limited to, at least one of the following: urgency level, accuracy requirements, etc. For example, if the meteorological decision-making task is related to drought disaster decisions in City A, then its location information is the location of City A, and the task type is data assessment of drought disaster. Accuracy requirements or other task characteristics can be directly written into the meteorological decision-making task, and some task characteristics can also be obtained through analysis based on the specific task situation. For example, in some examples, accuracy requirements can be determined based on the regional level (provincial, municipal, district) corresponding to the meteorological decision-making task; the determination method for each task characteristic can be predefined using a semantic analysis model or related determination rules.

[0054] It is evident that determining the task characteristics of a meteorological decision-making task can include directly obtaining information about the task characteristics carried within the meteorological decision-making task, or it can include obtaining the task characteristics of the meteorological decision-making task based on the description and analysis of the meteorological decision-making task.

[0055] S102. Based on location information and task characteristics, determine the target data range of the artificial intelligence model for the current meteorological decision-making task in a personalized manner;

[0056] It should be noted that the personalized pointer in this embodiment may result in different target data ranges for different geographical locations or different task characteristics. In some implementations of this embodiment, the target data range includes a location range and a time range described in relative terms. The location range within the target data range reflects the relationship with the geographical location corresponding to the meteorological decision-making task. For example, the target data range can describe a geographical radius R, indicating that meteorological data within the geographical radius R corresponding to the location of the meteorological decision-making task is required. The time range T within the target data range indicates that meteorological data within a time window T prior to the time required by the meteorological decision-making task (if not specifically specified, it is usually based on the current time of receiving the meteorological decision-making task). The target data range describes both the location and time ranges, thus enabling the location of meteorological data within a certain time period within a specific geographical area. The location and time ranges are described in relative terms, and only their relevant parameters (geographical radius R and time range T, etc.) need to be saved during use, making them easy to query and store. Of course, in practical applications, the location range can also be directly stored by defining the range using geographic coordinates or vector maps.

[0057] S103. Only meteorological data within the target data range are input into the corresponding artificial intelligence model to obtain meteorological decision support data;

[0058] Based on the determined target data range, corresponding meteorological data is loaded from the database. It is understood that the meteorological elements corresponding to the loaded meteorological data may differ depending on the meteorological decision-making task. For example, for decisions related to drought disasters, the loaded meteorological data may include precipitation and temperature, while for decisions related to sandstorm disasters, the meteorological elements may include wind speed and direction, and air pollution. However, in this embodiment, regardless of the meteorological elements, the input data will not exceed the target data range.

[0059] This embodiment determines the target data range individually based on location information and task characteristics, avoiding the need to import full or fixed-scale datasets into the AI ​​model for computation each time. Full or fixed-scale datasets typically contain a large amount of data that is not significantly helpful for meteorological decision-making. In this embodiment, because the target data range is individually determined based on the actual situation of the meteorological decision-making task, the effectiveness of the data acquired by the AI ​​model is higher in some implementation processes, helping to reduce resource consumption. The decision support data in this embodiment refers to data that can provide reference for subsequent disaster prevention, agricultural, and industrial decision-making processes; it is obtained by the AI ​​model in this embodiment based on the analysis of the input meteorological data. The AI ​​model in this embodiment is an AI model capable of providing data support and analysis for disaster or other decisions (such as agricultural irrigation) based on meteorological data. It can be an existing related model. The number of AI models can be arbitrary, i.e., one or more. Different AI models can be configured for different types of decision-making tasks.

[0060] In practical applications, the target data range can be determined through various optional methods to improve the effectiveness of meteorological data and minimize the input of invalid data.

[0061] As an example of implementation, the target data range can be dynamically determined based on the meteorological characteristics of the location corresponding to the meteorological decision-making task. For example, for locations with relatively complex surrounding terrain and meteorological conditions, the corresponding location range in the target data range can be relatively large; conversely, for locations with relatively simple surrounding terrain and meteorological conditions, the corresponding location range in the target data range can be relatively small. Alternatively, the time range in the target data range can be determined based on the rate of change of various meteorological elements at that location. In this embodiment, meteorological characteristics refer to meteorological features such as surrounding terrain and meteorological conditions. The system can also typically acquire geographic data, which can be used to determine the geographic and meteorological conditions of the area corresponding to the meteorological decision-making task.

[0062] As an example of implementation, the target data range can be determined based on at least one of the task characteristics of the meteorological decision-making task (e.g., urgency, accuracy requirements). For example, the target data range can be obtained by adjusting the base data range based on urgency and accuracy requirements. In practical applications, the base data range can be obtained based on experience and historical testing, and can be a range that can basically guarantee the effectiveness and reliability of meteorological decision support data.

[0063] The method described above can dynamically determine the target data range for a meteorological decision-making task based on its location information and / or task characteristics after receiving the task. The target data range can be calculated after receiving the meteorological decision-making task.

[0064] In other implementations, the target data range can be preset and stored. Upon receiving a meteorological decision-making task, the target data range matching the task is retrieved from the pre-stored data range information based on the task's location information and characteristics. In practice, the target data range can be manually set, but to ensure accuracy and reduce human workload, it can be automatically calculated using a specific calculation method.

[0065] As an example of implementation, the target data range can be determined based on the degree of contribution of meteorological data to decision-making. That is, meteorological data within the target data range are all meteorological data that contribute significantly to meteorological decision-making. The contribution of meteorological data to decision-making can be obtained based on the correlation analysis between historical meteorological data and historical event data. In some implementations, a pre-trained artificial intelligence model can be used to analyze the correlation between meteorological data and historical event data to determine the contribution of meteorological data to decision-making, and the data range containing meteorological data whose contribution to decision-making exceeds a set threshold can be used as the target data range. In this embodiment, historical event data refers to non-meteorological data related to decision-making or meteorological events, such as data related to meteorological disasters; for example, for floods, historical event data includes, but is not limited to, the scale of the flood, the peak height, and the arrival time; for droughts, historical event data includes, but is not limited to, the extent and duration of the drought.

[0066] In some implementations, historical meteorological data can be directly used for extrapolation to find corresponding artificial intelligence models that can effectively simulate a data range similar to the situation described by historical event data. It is understood that the geographical characteristics and meteorological features of the same location typically do not change significantly over time; therefore, the data range determined based on historical meteorological data is also applicable to future decision-making. These implementations do not require the configuration of additional data correlation analysis models, resulting in a simple system; furthermore, their results have a high degree of compatibility with the currently used artificial intelligence models, eliminating errors caused by differences between different models.

[0067] The target data range determined by the methods exemplified above is primarily related to geographical characteristics and the characteristics of artificial intelligence models. Therefore, it is more suitable for pre-calculating and storing appropriate target data ranges, which can then be directly retrieved through querying. In the methods exemplified above, a large number of data ranges are typically stored, each corresponding to different geographical locations and task requirements. For ease of distinction, in the embodiments of this invention, these pre-stored data ranges are uniformly referred to as candidate data ranges; while the target data range can be understood as the data range determined in a specific meteorological decision-making process. The methods exemplified above can determine and store candidate data ranges before receiving a meteorological decision-making task. After receiving the task, the corresponding candidate data range is retrieved from the stored results based on location information and / or task characteristics as the target data range. Furthermore, the methods exemplified above determine candidate data ranges based on historical meteorological data. These candidate data ranges can be calculated during periods of idle computing power, avoiding the need to consume large amounts of computing power in a short period.

[0068] In some implementations, inputting only meteorological data within the target data range into the corresponding artificial intelligence model to obtain meteorological decision support data includes:

[0069] Based on the accuracy requirements of the meteorological decision-making task, the meteorological data within the target data range is completely input into the corresponding artificial intelligence model, or, based on the accuracy requirements, the meteorological data within the target data range is sampled and input into the corresponding artificial intelligence model.

[0070] In some examples, accuracy requirements can influence the determination of the target data range and, consequently, the amount of data read from that range. When accuracy requirements are high, all relevant meteorological data within the target data range is input into the AI ​​model. Conversely, when accuracy requirements are low, only a portion of the data may be input. In this embodiment, for cases with low accuracy requirements, the input data is selected through data sampling. Data sampling in this embodiment refers to selecting meteorological data at specific step sizes. For example, the original meteorological data may include precipitation data recorded every five minutes, but when accuracy requirements are low, precipitation data within the target data range is read in 30-minute intervals. The specific sampling interval is set according to the accuracy requirements. Selecting whether to input all meteorological data within the target data range based on accuracy requirements can further reduce the amount of data input to the AI ​​model in some implementations, thereby further reducing computational demands. Example

[0071] This embodiment illustrates the process of determining the target data range using a relatively specific example. For steps not mentioned in this embodiment, please refer to other embodiments of the present invention; where there is no conflict, the steps exemplified in the various embodiments can be combined.

[0072] See Figure 2 As shown, in this embodiment, before individually determining the target data range of the artificial intelligence model for the current meteorological decision-making task, the following steps are included, but are not limited to:

[0073] S201. During periods of idle computing power, multiple event simulations are performed based on historical event data to obtain simulation data.

[0074] In this embodiment, historical meteorological data of different ranges are input into the artificial intelligence model for each event simulation. This embodiment directly uses the artificial intelligence model to simulate historical meteorological data, attempting to reconstruct simulation results that closely resemble the historical event data. Simulations performed in this manner can identify the data range most suitable for the artificial intelligence model.

[0075] Understandably, multiple event simulations can be performed at different times, i.e., they can be conducted intermittently. The data range differs in each simulation, including the location range and / or the time range. The data range can be adjusted in increments, such as increasing the geographical radius by 1 km until the upper limit, or increasing the time range by 1 day until the upper limit.

[0076] Extrapolated data is data obtained by extrapolating similar events based on historical meteorological data. For example, taking the flood in City A as a historical event, the historical event data includes the flood's scale, peak height, and arrival time; correspondingly, the extrapolated data includes the simulated flood's scale, peak height, and arrival time. The specific elements included in the historical event data depend on what is recorded in the historical data and what the artificial intelligence model used is capable of analyzing.

[0077] Each event simulation inputs historical meteorological data of different ranges into the artificial intelligence model.

[0078] Suppose that the first simulation inputs historical meteorological data, including rainfall, hydrological data, and temperature, for the area within city A one week prior to the flood. The artificial intelligence model can then use this historical meteorological data to simulate the flood and obtain a simulated flood result. For example, it might obtain simulation data including the scale of the flood, the peak height, and the arrival time.

[0079] The second simulation can use a different data range, for example, adding data from surrounding provinces and cities in addition to the area under the jurisdiction of City A. If the time range remains the same, historical meteorological data such as precipitation, hydrological data, and temperature from the week before the flood are input for City A and its surrounding provinces and cities. The artificial intelligence model can then simulate the flood based on this historical meteorological data, obtaining another simulated flood result, including simulated flood scale, simulated peak height, simulated arrival time, and other extrapolated data.

[0080] The third simulation can use different data ranges, such as expanding or narrowing the time range. For example, input historical meteorological data such as precipitation, hydrological data, and temperature for the two weeks prior to the flood within the jurisdiction of City A.

[0081] This allows for multiple simulations, with the data range adjusted each time to obtain multiple sets of simulation data.

[0082] S202. Compare the simulation data with historical event data;

[0083] In this embodiment, the acquired historical meteorological data and historical event data are corresponding, meaning that the historical meteorological data and historical event data used in the simulation correspond to the same location and the same time range. The steps in this embodiment are used to calculate and store the candidate data range for a certain location.

[0084] Specifically, the comparison can be to calculate the degree of similarity between the two, or to determine whether the inferred event is comparable in scale to the historical event.

[0085] Each simulation yields data such as the simulated flood size, peak height, and arrival time. However, this simulation is based on historical meteorological data, which actually includes historical data on the size, peak height, and arrival time of real floods. In the example above, the simulated flood size, peak height, and arrival time are compared with historical data on the size, peak height, and arrival time of real floods. This allows us to determine which simulation result is closer to reality and thus identify the optimal data range for that simulation. Simulated data and historical event data can be directly compared (e.g., the simulated flood size, peak height, and arrival time can be directly compared with the actual flood size, peak height, and arrival time). An accuracy can be calculated for each specific data item (e.g., the error between the simulated peak height and the actual peak height), and a final similarity score is obtained by combining the accuracy scores of each data point. S203. Determine the candidate data range based on the comparison results between the simulated data and historical event data;

[0086] In this embodiment, the minimum data range where the similarity reaches a set threshold can be used as the candidate data range; or the minimum data range that can be used to extrapolate an event of comparable scale to historical events can be used as the candidate data range. The candidate data range determined by comparing the extrapolated data with historical event data is the data range within which the artificial intelligence model can obtain relatively accurate results, thus enabling the final meteorological decision support data of this embodiment to have high reliability.

[0087] By determining which projection results are closer to reality, we can identify the optimal data range used in that projection. This data range can then be recorded as a candidate range, which may be selected for future meteorological decision-making tasks.

[0088] In the example above, assuming the results of the second simulation have the highest degree of agreement with historical event data, the candidate data range is "location range: City A and its surrounding provinces and cities; time range: one week". Meteorological elements such as precipitation, hydrological data, and temperature are determined based on the decision-making needs of flood control. That is, when it is known that the meteorological decision-making task is flood control, it can be determined that the data of these meteorological elements need to be retrieved, therefore, this content does not need to be stored.

[0089] S204. Associate and store the geographic information and features corresponding to the candidate data range with the historical event data.

[0090] For example, the data can be stored in a lookup table. Users can directly query the table using the location information and task characteristics corresponding to the meteorological decision-making task. If the data matches the location information and characteristics of a historical event in the lookup table, the corresponding candidate data range can be determined as the target data range. Of course, in practical applications, the representation and storage method of the data range can be arbitrary.

[0091] For example, a meteorological decision-making task corresponds to area B, and the task type is flood disaster prediction and assessment. As an example, suppose the target data range for flood disaster prediction in area B, retrieved from a lookup table, is a geographical radius of 50 km and spans 10 days. Therefore, the data input into the artificial intelligence model is meteorological data related to floods within a 50 km radius centered on area B, spanning the 10 days prior to the current time.

[0092] In the example, the candidate data range is defined as "City A and its surrounding provinces and cities (location range), one week (time range)". This is then associated with the corresponding geographic information "City A" and the feature "flood" or "flood control" and stored accordingly. The format for determining the data range is similar for other disasters such as droughts and sandstorms, or other meteorological events.

[0093] In some implementations, if the system does not yet store a matching alternative data range, the target data range can be determined through other dynamically calculated methods. For example, adjustments can be made based on the urgency and accuracy requirements of the basic data range to obtain the target data range. When computing power is idle, the above-described method in this embodiment, which uses an artificial intelligence model to calculate a more suitable target data range, can be continuously calculated and updated in the lookup table.

[0094] Understandably, this is to ensure a balance between computational resources consumed and accuracy in the meteorological decision-making process. When a larger data range fails to effectively improve the accuracy of the extrapolation, the extrapolation can be stopped, and the final data range can be used as the target data range.

[0095] As an example, such as Figure 3 As shown, multiple event simulations are performed based on historical event data to obtain simulation data. Each simulation inputs historical meteorological data of different ranges into the artificial intelligence model, including but not limited to the following steps:

[0096] S301. Starting from the basic data range, continuously expand the data range in event simulation;

[0097] Based on the comparison between the projected data and historical event data, the range of candidate data is determined to include:

[0098] S302. Based on the trend of similarity changes between each simulation data and historical event data, determine the optimal data range as the candidate data range;

[0099] In this embodiment, the optimal standard can be defined by relevant professionals according to their needs. For example, the following steps can be used to analyze the trend of similarity changes and determine the optimal data range.

[0100] S3021. Determine several data ranges where the similarity between the inferred data and historical event data is greater than the minimum similarity threshold, and sort these data ranges from low to high according to the amount of data.

[0101] S3022. Sequentially determine whether the increase in similarity between the inferred data and historical event data is less than the minimum increase threshold after each increase in data volume. When it is determined that the increase in similarity between the N+1th inferred data and historical event data compared to the Nth time is less than the minimum increase threshold, the data range used in the Nth inferred data is taken as the candidate data range.

[0102] As another example, an evaluation decay coefficient can be set based on the amount of data; the larger the amount of data, the larger the evaluation decay coefficient. When the increase in similarity between the inferred data and historical event data resulting from expanding the data range is less than the evaluation decay coefficient, the data range is no longer expanded, and the final data range is used as the alternative data range.

[0103] As can be seen, this embodiment can find the optimal data range within the specified range through various methods, achieving a balance between the amount of data in the final determined target data range and the accuracy for meteorological decision-making as required by the user. This avoids inputting a large amount of data that provides little or no benefit to accuracy, ensuring the effectiveness of the data input to the artificial intelligence model.

[0104] For example, the similarity between the inferred data and historical event data in this embodiment can be determined by comparing their temporal and spatial consistency, or by comparing the similarity of their key features. For temporal consistency, the Segmented Dynamic Time Warping (SDTW) algorithm can be used to calculate the Dynamic Time Warping (DTW) distance; spatial consistency can be calculated using algorithms including, but not limited to, Hausdorff distance (HD). Key feature similarity can be calculated by constructing an event feature fingerprint vector F and using cosine similarity to calculate the similarity between the inferred and real fingerprint vectors. In practical applications, the similarities of the above different dimensions can be summarized using a weighted combination to obtain the final similarity result. The above only exemplifies a few optional similarity calculation methods; in practical applications, any method can be used, and this invention is not limited. In this embodiment, regardless of the comparison method for similarity, a relatively optimal data range can be found based on the relative accuracy and trends of different data ranges during event inference.

[0105] The method described in this embodiment is based on historical meteorological data and compared with real historical events. The resulting target data range is more closely aligned with the characteristics of the artificial intelligence model and the actual conditions of the corresponding region, resulting in high accuracy. Furthermore, the method described in this embodiment can execute at least some steps when computing power is idle, reducing sudden increases in computing power consumption and improving the utilization rate of computing resources. Example

[0106] This embodiment illustrates the process of determining the target data range by another example. For steps not mentioned in this embodiment, please refer to the other embodiments of the present invention. Unless otherwise specified, the steps illustrated in the embodiments can be combined.

[0107] In this embodiment, the target data range is determined primarily based on the influencing factors in the task characteristics, enabling rapid dynamic computation with minimal computational resource consumption. In this embodiment, the task characteristics of the meteorological decision-making task include, but are not limited to, urgency and accuracy requirements. For meteorological decision-making tasks with high urgency, the final determined target data range can be smaller to accelerate the output speed of the artificial intelligence model; for meteorological decision-making tasks with low urgency, the target data range can be appropriately expanded to obtain more accurate meteorological decision support data. Conversely, for meteorological decision-making tasks with high accuracy requirements, the final determined target data range can be larger to provide more data; for meteorological decision-making tasks with low accuracy requirements, the final determined target data range can be smaller. It is understood that the terms "larger" and "smaller" in this embodiment refer to relative degrees, that is, relative magnitudes assuming other variables are equal.

[0108] In this embodiment, the factors affecting the target data range are referred to as influencing factors. For example, when determining the target data range for a meteorological decision-making task, the influencing factors related to the target data range in the meteorological decision-making task are identified. In this example, the influencing factors include, but are not limited to, at least one of urgency, accuracy requirements, terrain complexity, and meteorological complexity. In practical applications, more types of influencing factors can be introduced, and their application methods are the same as those exemplified later.

[0109] In this embodiment, the influence factor is used to determine the location range within the target data range. See also: Figure 4 In this embodiment, the method for determining the target data range may include:

[0110] S401. Obtain the basic location range for meteorological decision-making tasks;

[0111] S402. Determine the influencing factors of the location range and the corresponding influence coefficients of each influencing factor;

[0112] S403. Multiply the basic location range by each influence coefficient to obtain the location range in the target data range;

[0113] Taking urgency and accuracy requirements as examples, these two influencing factors are usually directly incorporated into meteorological decision-making tasks. In practical applications, urgency and accuracy requirements can also be obtained through analysis of other information within the meteorological decision-making task. Assuming the urgency coefficient P and accuracy requirement coefficient Q of the meteorological decision-making task are determined, the target location range R is determined according to the following formula 1, where R_base is the base location range, and α and β are the adjustment coefficients for urgency and accuracy requirements, respectively.

[0114] Formula 1: R = R_base × (1 + α×P) × (1 + β×Q)

[0115] In this embodiment, the base location range can be 50-300km, which can be determined based on historical results. The adjustment coefficient α for urgency can range from 0.1 to 0.5, which can be determined based on historical event analysis. The adjustment coefficient β for accuracy requirement can range from 0.2 to 0.6, which can be determined based on the analysis of the relationship between error and data volume. The urgency coefficient P and the accuracy requirement coefficient Q range from 0 to 1. An urgency coefficient P of 0 indicates a normal level of urgency, while a value of 1 indicates the highest level of urgency. Similarly, an accuracy requirement coefficient Q of 0 indicates the lowest accuracy requirement, while a value of 1 indicates the highest accuracy requirement. The urgency coefficient P and the accuracy requirement coefficient Q can be quantitatively set by relevant personnel according to actual needs. For example, the urgency of a decision-making task related to agricultural irrigation can be 0.3, and the accuracy requirement can be configured to 0.6; the urgency of a decision-making task related to severe disaster prevention and control can be 1, and the accuracy requirement can be configured to 0.8 or even 1.

[0116] Assuming that in the decision analysis of the impact of a typhoon event, the base location range R_base is 150 km, the adjustment coefficient for urgency α is 0.3, the adjustment coefficient for accuracy requirement β is 0.25, the urgency coefficient P is 0.9, and the accuracy requirement coefficient Q is 0.8, then the final determined location range R is 150 × (1 + 0.3 × 0.9) × (1 + 0.25 × 0.8) = 228.6 km. In practical applications, upper and lower limits can be set for the location range. If the calculated result exceeds the upper or lower limit, the upper or lower limit value is taken. The time range's impact on the accuracy of meteorological decision-making is more easily affected by the rate of change of meteorological elements. That is, when the rate of change of meteorological elements is fast, meteorological data over a longer time range provides less reference value; when the rate of change of meteorological elements is slow, meteorological data over a shorter time range cannot provide sufficient data for reference. Therefore, the time range within the target data range can be determined independently of the location range. Specifically, in this embodiment, the time range can be determined based on the rate of change of meteorological elements. For example, a base time range is also set. In practical applications, the base time range may differ for different meteorological elements. When determining the target data range, the rate of change of each meteorological element corresponding to the current meteorological decision-making task is determined. Based on the base time range, the target time range is determined according to the rate of change of the meteorological elements. The length of the time range is negatively correlated with the rate of change of the meteorological elements. It is understandable that the target time range may differ for different meteorological elements. In practical applications, to facilitate data processing, the maximum time range calculated for each meteorological element can be selected as the target time range for this meteorological decision-making task, ensuring data time alignment. The target time range and the target location range together determine the target data range corresponding to the meteorological decision-making task.

[0117] As an example, the base time range for some meteorological elements can be set as follows: temperature - 48 hours; wind speed - 24 hours; precipitation - 12 hours. Depending on the rate of change of the meteorological element, the time range can be between 50% and 200% of the base time range. For example, if the current rate of change in precipitation is 10% higher than the baseline rate of change, then the time range can be 110% of the base time range. In practical applications, the correspondence between the rate of change and the base time range, as well as the specific range, can be set by technical personnel according to requirements. Example

[0118] This embodiment illustrates the process of determining the target data range using another example. This example combines the target data range determination methods exemplified in Embodiments Two and Three above, such as... Figure 5 As shown, in this embodiment, after receiving the meteorological decision-making task, the following steps are performed, including but not limited to:

[0119] S501. Based on the location information and task characteristics of the meteorological decision-making task, query from the pre-stored data range information whether there is a candidate data range that matches the meteorological decision-making task.

[0120] In this embodiment, the range of candidate data stored is determined by extrapolating from historical meteorological data and comparing it with historical events, as illustrated in Embodiment 2 above.

[0121] S502. After finding a range of matching candidate data stored in the database, determine it as the target data range.

[0122] S503. If no matching alternative data range is found, the target data range of the meteorological decision-making task is dynamically calculated based on the meteorological characteristics of the location corresponding to the meteorological decision-making task and / or the task characteristics.

[0123] The process of dynamically calculating the target data range based on the task characteristics of meteorological decision-making can be seen in the example of Embodiment 3 of the present invention.

[0124] In this embodiment, the target data range derived from historical meteorological data using an artificial intelligence model is preferentially selected for meteorological decision-making, resulting in high accuracy. Even when no target data range is pre-stored, dynamic calculation can be used to determine the target data range, allowing for flexible adaptation to real-world scenarios. For various meteorological decision-making tasks, a relatively reasonable target data range can be determined individually, avoiding the need to import large amounts of potentially invalid data into the artificial intelligence model. Furthermore, in practical applications, when computing power is idle, missing parameters for the target data range in the system can be continuously calculated and supplemented. Over a certain period, the target data range in the system can gradually become more comprehensive. Example

[0125] This embodiment provides a meteorological decision support and optimization system 100 based on artificial intelligence. See [link to documentation]. Figure 6 The system includes a task parsing unit 101, a data range determination unit 102, a data transmission unit 103, and several artificial intelligence models 104.

[0126] The task parsing unit 101 is used to determine the location information and task characteristics corresponding to the meteorological decision-making task after receiving the meteorological decision-making task.

[0127] The data range determination unit 102 is used to determine the target data range of the artificial intelligence model for the current meteorological decision-making task in a personalized manner based on location information and task characteristics.

[0128] The data transmission unit 103 is used to input only the meteorological data within the target data range into the corresponding artificial intelligence model 104.

[0129] Several artificial intelligence models 104 are used to determine meteorological decision support data based on the input meteorological data. In this embodiment, the number of artificial intelligence models 104 can be arbitrary, that is, there can be one or more. It is understood that different artificial intelligence models 104 can be configured for different types of meteorological decision-making tasks. The target data range determined in this embodiment also corresponds to the artificial intelligence model 104. In practical applications, since there are differences between artificial intelligence models 104, different target data ranges may be used when applying different artificial intelligence models 104 for the same meteorological decision-making task.

[0130] The specific steps that the units in the AI-based meteorological decision support and optimization system 100 of this embodiment can execute can also be referred to the descriptions of the above embodiments, and will not be repeated in this embodiment.

[0131] Furthermore, although exemplary embodiments have been described herein, their scope includes any and all embodiments based on this disclosure that have equivalents, modifications, omissions, combinations (e.g., schemes overlapping various embodiments), adaptations, or changes. The examples described are not limited to those described in this specification or during the implementation of the invention and are to be interpreted as non-exclusive.

[0132] The above description is intended to be illustrative and not restrictive. For example, the above examples (or one or more of them) can be used in combination with each other. Other embodiments can be used by those skilled in the art when reading the above description.

[0133] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, various modifications and variations can be made to the embodiments of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A meteorological decision support and optimization method based on artificial intelligence, characterized in that, include: Upon receiving a meteorological decision-making task, the location information and task characteristics corresponding to the meteorological decision-making task are determined; Based on the location information and task characteristics of the meteorological decision-making task, a query is performed from the pre-stored data range information to determine whether there is a candidate data range that matches the meteorological decision-making task; the candidate data range is a data range determined by the artificial intelligence model using historical meteorological data for extrapolation. After failing to find a matching candidate data range, the target data range for the meteorological decision-making task is dynamically calculated based on the meteorological characteristics of the location corresponding to the task and / or the task characteristics. The dynamic calculation of the target data range includes: obtaining the basic location range of the meteorological decision-making task; determining the influencing factors of the location range and the corresponding influence coefficients of each influencing factor; multiplying the basic location range by each influence coefficient to obtain the location range within the target data range; and obtaining the basic time range of the meteorological decision-making task; determining the rate of change of the meteorological elements corresponding to the meteorological decision-making task; and, based on the basic time range, determining the time range within the target data range based on the rate of change of the meteorological elements, wherein the length of the time range is negatively correlated with the rate of change of the meteorological elements; the influencing factors include at least one of urgency, accuracy requirements, terrain complexity, and meteorological complexity. Meteorological data within the target data range is input into the corresponding artificial intelligence model to obtain meteorological decision support data.

2. The meteorological decision support and optimization method based on artificial intelligence as described in claim 1, characterized in that, Before querying the pre-stored data range information for a matching alternative data range based on the location information and task characteristics of the meteorological decision-making task, the method further includes: During periods of idle computing power, multiple event simulations are performed based on historical event data to obtain simulation data. Each event simulation inputs historical meteorological data of different data ranges into the artificial intelligence model. The inferred data is compared with the historical event data; The range of candidate data is determined based on the comparison results between the inferred data and the historical event data; The candidate data range is associated with and stored in relation to the location information and features corresponding to historical event data.

3. The meteorological decision support and optimization method based on artificial intelligence as described in claim 2, characterized in that, The process of obtaining the extrapolated data by performing multiple event extrapolations based on the historical event data includes: starting from the basic data range and continuously expanding the data range in the event extrapolation; The step of determining the candidate data range based on the comparison results between the inferred data and the historical event data includes: determining the optimal data range as the candidate data range based on the trend of similarity changes between each inferred data and the historical event data.

4. The meteorological decision support and optimization method based on artificial intelligence as described in claim 3, characterized in that, The step of determining the optimal data range as the candidate data range based on the similarity trend between each simulation data and the historical event data includes: Several data ranges are identified where the similarity between the inferred data and the historical event data is greater than the minimum similarity threshold, and these data ranges are sorted from low to high according to the amount of data. After each increase in data volume, it is determined whether the increase in similarity between the inferred data and the historical event data is less than the minimum increase threshold. When it is determined that the increase in similarity between the (N+1)th inferred data and the historical event data compared to the Nth time is less than the minimum increase threshold, the data range used in the Nth inferred data is taken as the candidate data range.

5. The meteorological decision support and optimization method based on artificial intelligence as described in claim 1, characterized in that, After finding a matching candidate data range stored, the matching candidate data range is determined as the target data range.

6. The meteorological decision support and optimization method based on artificial intelligence as described in any one of claims 1 to 5, characterized in that, The target data range includes a location range and a time range described in relative terms; The location range indicates the geographic radius relative to the location information; The time range indicates the time span relative to the time corresponding to the meteorological decision-making task.

7. A meteorological decision support and optimization system based on artificial intelligence, characterized in that, include: The task parsing unit is used to determine the location information and task characteristics corresponding to the meteorological decision-making task after receiving the meteorological decision-making task. A data range determination unit is used to query, based on the location information and task characteristics of the meteorological decision-making task, whether there are any candidate data ranges matching the meteorological decision-making task from pre-stored data range information; the candidate data ranges are data ranges determined by an artificial intelligence model using historical meteorological data; if no matching candidate data ranges are found, the unit dynamically calculates the target data range of the meteorological decision-making task based on the meteorological characteristics of the location corresponding to the meteorological decision-making task and / or the task characteristics; dynamically calculating the target data range of the meteorological decision-making task includes: obtaining the basic location range of the meteorological decision-making task; determining the influencing factors of the location range and the influence coefficients corresponding to each influencing factor; multiplying the basic location range by each of the influence coefficients to obtain the location range in the target data range; and obtaining the basic time range of the meteorological decision-making task; determining the rate of change of the meteorological elements corresponding to the meteorological decision-making task; and, based on the basic time range, determining the time range in the target data range based on the rate of change of the meteorological elements, wherein the length of the time range is negatively correlated with the rate of change of the meteorological elements; the influencing factors include at least one of urgency, accuracy requirements, terrain complexity, and meteorological complexity. The data transmission unit is used to input only the meteorological data within the target data range into the corresponding artificial intelligence model; Several artificial intelligence models are used to determine meteorological decision support data based on the input meteorological data.

Images