Meteorological intelligent service system based on language large model and construction method

Abstract

The application discloses a meteorological intelligent service system and a construction method based on a language large model, and belongs to the technical field of language large models. The system comprises modules of meteorological data integration, meteorological service business construction and demand analysis, service matching degree record and response set update, and meteorological knowledge graph construction and personalized portrait generation; and the construction method sequentially realizes meteorological data interface set formation, structured meteorological service business layer construction and demand analysis, node service response set dynamic update, meteorological knowledge graph and personalized portrait generation. The application realizes accurate demand and scene adaptation by unifying multi-source meteorological data coding and interface grouping and combining a language large model, introduces time limit and control update service matching degree, and constructs a knowledge graph associated with service nodes, so that the pain points of scattered meteorological service data sources, low demand matching accuracy, no time limit for service, and lack of personalization are solved, and efficient, accurate and personalized meteorological services in agricultural, traffic, tourism and other scenes can be realized.

Description

Technical Field

[0001] This invention relates to the field of language large model technology, specifically to a meteorological intelligent service system and its construction method based on a language large model. Background Technology

[0002] Existing meteorological data comes from the National Meteorological Center, provincial monitoring stations, and industry-specific equipment (such as agricultural soil monitoring instruments and traffic visibility sensors). The interface formats (JSON / XML) and transmission protocols (HTTP / MQTT) of these data sources differ significantly. When the system calls the data, it needs to develop adaptation programs for different interfaces. For example, when calling "precipitation data (National Meteorological Center, XML format) + soil moisture data (agricultural monitoring station, JSON format)," the traditional system needs to adapt to the two types of interfaces separately. Moreover, the high error rate of data transmission caused by interface incompatibility seriously affects the service response efficiency.

[0003] Furthermore, traditional meteorological services rely on "keyword matching" to analyze needs, which cannot be linked to specific application scenarios. For example, if a user asks "What should I bring for hiking in Lushan next week?", the existing system only returns "30% probability of precipitation in the next 7 days" for the Lushan area, without considering the specific needs of "mountain hiking" regarding wind force (affecting hiking safety) and ultraviolet radiation (affecting sun protection). Similarly, if a farmer asks "Does spring plowing need irrigation?", the system only returns precipitation data, without linking the adaptation relationship between "spring crops (wheat / rice) + soil moisture", resulting in low accuracy in demand matching and insufficient service practicality.

[0004] Secondly, meteorological services are highly time-sensitive (e.g., typhoon warnings are 24 hours away, and agricultural temperature monitoring is 7 days away), but existing systems often use outdated data, leading to service deviations. For example, in typhoon warnings, the system still uses wind speed data from 24 hours ago, resulting in a deviation of more than 3 levels between the warning result and the actual wind force. In agricultural spring plowing scenarios, using soil moisture data from one week ago to guide irrigation can easily cause crop waterlogging or drought, and the accuracy of disaster emergency response is insufficient.

[0005] Furthermore, existing systems mostly provide "regionally unified" meteorological information, which cannot generate personalized services based on user identity (farmer / traveler / logistics driver) and historical needs. At the same time, the various meteorological service functions (precipitation warning, temperature monitoring, humidity advice) are independent of each other and do not form a service node connection, so they cannot provide comprehensive advice (such as "low temperature + high soil moisture during the wheat greening period, cold and flood prevention measures are required"), making it difficult to meet the in-depth needs of vertical fields.

[0006] Furthermore, existing meteorological services cannot adapt to the personalized needs of diverse scenarios, and there is an urgent need for an intelligent service solution that can achieve full-link optimization of "data integration - demand adaptation - timeliness control - personalized output". Summary of the Invention

[0007] The purpose of this invention is to provide a meteorological intelligent service system and its construction method based on a large language model, so as to solve the problems mentioned in the background art.

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0009] A meteorological intelligent service system based on a large language model includes: a meteorological data integration module, a meteorological service business construction and demand analysis module, a service matching degree recording and response set update module, and a meteorological knowledge graph construction and personalized profile generation module.

[0010] The meteorological data integration module is used to coordinate meteorological data sources, standardized data interfaces, and data transmission channels to form a data interface set.

[0011] The meteorological service business construction and demand parsing module is used to construct a structured meteorological service business layer and parse user service demands to call the data transmission channel;

[0012] The service matching degree recording and response set update module is used to obtain the service matching degree and update the node service response set;

[0013] The meteorological knowledge graph construction and personalized profile generation module is used to quantify the scope of adaptation and association, evaluate the service interaction value and the degree of integration of personalized services, and form a meteorological knowledge graph and a personalized meteorological service profile.

[0014] As a preferred embodiment of the present invention, the meteorological data integration module includes a data source and interface coordination unit and a data interface set construction unit;

[0015] The data source and interface coordination unit is used to coordinate all meteorological data sources within the meteorological service system, the standardized data interfaces corresponding to each meteorological data source, and the data transmission channels associated with each standardized data interface, and to uniformly encode the meteorological data sources, standardized data interfaces, and data transmission channels respectively.

[0016] The data interface set construction unit is used to group each standardized data interface through the data transmission channel to form a data interface set corresponding to the data transmission channel, wherein the data transmission channel includes standardized data interfaces that provide data support for meteorological services.

[0017] As a preferred embodiment of the present invention, the meteorological service business construction and demand analysis module includes a structured business layer construction unit and a demand analysis and channel invocation unit;

[0018] The structured business layer construction unit is used to construct a structured meteorological service business layer based on the service requirement parsing capability of the language big model and the meteorological service scenario. One service requirement corresponds to one meteorological service scenario to form a structured meteorological service business layer, and each structured meteorological service business layer is used as a language big model node.

[0019] The requirement parsing and channel invocation unit is used to parse user service requirements and invoke corresponding data transmission channels based on the service requirement parsing capability of the language big model node and the weather service scenario adaptation function. At the same time, it captures the invoked data transmission channels to realize the instruction closed loop of service requirement parsing of the language big model and weather service scenario adaptation.

[0020] As a preferred embodiment of the present invention, the service matching degree recording and response set update module includes a service matching degree acquisition unit and a response set generation and update unit;

[0021] The service matching degree acquisition unit is used to acquire the service matching degree of the data transmission channel in the process of demand analysis and meteorological service scenario adaptation when the language large model node calls the data transmission channel. The service matching degree is the adaptation ratio of the meteorological data source, that is, the ratio of the number of standardized data interfaces corresponding to the meteorological data source adapted to the meteorological service scenario in the data interface set to the total number of standardized data interfaces in the data interface set.

[0022] The response set generation and update unit is used to collect the service matching degree generated when the language large model node calls each data transmission channel and generate the node service response set. At the same time, it pre-constructs the applicable time range segment of the meteorological service scenario, uses this time range segment as the response time range of each structured meteorological service business layer, and updates the node service response set within this response time range.

[0023] As a preferred embodiment of the present invention, the meteorological knowledge graph construction and personalized profile generation module includes an adaptation association range quantification unit, a service interaction and integration degree evaluation unit, and a knowledge graph and profile generation unit.

[0024] The adaptation association range quantization unit is used to select the maximum and minimum values ​​of service matching degree in the updated node service response set, respectively, so as to quantify the upper and lower limits of the adaptation association range of each language large model node.

[0025] The service interaction and integration evaluation unit is used to evaluate the unstructured service interaction value between the structured meteorological service business layers corresponding to different language large model nodes based on the adaptation association range, and evaluate the personalized service integration degree based on the unstructured service interaction value.

[0026] The knowledge graph and profile generation unit is used to preset a personalized service fusion threshold. If the personalized service fusion degree is greater than or equal to the threshold, a connection edge is constructed between the corresponding two language large model nodes. Based on the construction results of all connection edges, a meteorological knowledge graph of the language large model nodes is formed. This meteorological knowledge graph is the user's personalized meteorological service profile.

[0027] A method for constructing intelligent meteorological services based on a large language model, comprising the following steps:

[0028] Step S1: Using structured data, meteorological data sources within the meteorological service system, standardized data interfaces corresponding to each meteorological data source, and data transmission channels associated with each standardized data interface are formed. The standardized data interfaces are then grouped through the data transmission channels to form a data interface set for the data transmission channels.

[0029] Step S2: Based on the service requirement analysis of the language big model and the meteorological service scenario, construct a structured meteorological service business layer between the service requirements and the meteorological service scenario, which serves as a node of the language big model, and parse the user's service requirements through the language big model node to call the data transmission channel;

[0030] Step S3: Record the service matching degree of the data transmission channel called in the process of demand analysis and meteorological service scenario adaptation to the node service response set, and update the service response set of each node based on the applicable time range of the meteorological service scenario;

[0031] Step S4: Based on the updated node service response set, quantify the adaptation and association range of the language big model nodes, and evaluate the unstructured service interaction value and personalized service integration degree between the structured meteorological service business layers corresponding to different language big model nodes, so as to build connection edges between language big model nodes and form a meteorological knowledge graph of language big model nodes. This meteorological knowledge graph serves as a personalized meteorological service profile for users.

[0032] As a preferred embodiment of the present invention, the specific implementation process of step S1 includes:

[0033] The system coordinates all meteorological data sources within the meteorological service system, as well as the standardized data interfaces corresponding to each meteorological data source and the data transmission channels associated with each standardized data interface, and uniformly encodes the meteorological data sources, standardized data interfaces, and data transmission channels respectively;

[0034] The data transmission channel includes standardized data interfaces that provide data support for meteorological services. These standardized data interfaces are grouped together through the data transmission channel to form a data interface set for the data transmission channel.

[0035] As a preferred embodiment of the present invention, the specific implementation process of step S2 includes:

[0036] Based on the service demand analysis of the language big model and the meteorological service scenario, a structured meteorological service business layer is constructed. In this layer, a service demand and a meteorological service scenario constitute a structured meteorological service business layer, and a structured meteorological service business layer is used as a node of the language big model.

[0037] Based on the service requirement parsing capability and meteorological service scenario adaptation function of the language big model node, each data transmission channel is captured. If a data transmission channel is called after parsing the user's service requirement through the language big model node, the called data transmission channel is captured to realize the instruction closed loop of service requirement parsing of the language big model and meteorological service scenario adaptation.

[0038] As a preferred embodiment of the present invention, the specific implementation process of step S3 includes:

[0039] Based on the capture behavior of the data transmission channel, when the data transmission channel is called through the language large model node, the service matching degree of the called data transmission channel in the process of demand analysis and meteorological service scenario adaptation is obtained. The service matching degree refers to the adaptation ratio of the meteorological data source. The numerator of the adaptation ratio is the number of standardized data interfaces corresponding to the meteorological data source adapted to the meteorological service scenario in the data interface set, and the denominator of the adaptation ratio is the total number of standardized data interfaces included in the data interface set.

[0040] Collect the service matching degree generated when the language large model node calls each data transmission channel, and generate a node service response set; pre-construct the applicable time range segment of the meteorological service scenario, and use the time range segment as the response time range of each structured meteorological service business layer, and update the service response set of each node within the response time range.

[0041] As a preferred embodiment of the present invention, the specific implementation process of step S4 includes:

[0042] In the updated service node set, the maximum and minimum service matching scores are selected respectively to quantify the upper and lower limits of the adaptation and association range of the language large model nodes:

[0043] upper limit ;

[0044] lower limit ;

[0045] In the formula, This represents the e-th language large model node. This represents the language large model node within the x-th response time range. The corresponding updated node service response set, and , This represents the r-th data transmission channel. Representation of language large model nodes Invoke data transmission channel The service matching degree generated at that time and These are, in order, the maximum value function and the minimum value function. This represents the mean of the service matching score. For the variance of service matching degree, and , R represents the total number of data transmission channels, and E represents the total number of language large model nodes;

[0046] Then the language large model nodes The adaptation association range within the x-th response timeframe is ;

[0047] It should be noted that service matching degree is a quantitative indicator of the degree of fit between a "language large model node (corresponding to a service requirement + scenario)" and a "data channel" (for example, the node "wheat greening period low temperature warning" calls the "temperature + precipitation channel", the number of temperature interfaces that fits is 1, the total number of interfaces is 2, and the matching degree = 0.5). The maximum value of the service matching degree represents the "optimal fit state" that the node can achieve within the current time range (such as 7 days of agricultural monitoring time), and the minimum value of the service matching degree represents the "worst fit state" of the node. The mean (average of all matching degrees) and variance (dispersion of matching degree) are used to "standardize" the upper and lower limits, avoid the bias of a single extreme value, and ensure that the range can reflect the stable fit capability of the node (rather than accidental cases), thereby transforming the discrete matching degree into a "continuous range based on statistical distribution", thus accurately describing the boundary of the node's fit capability.

[0048] Based on the scope of adaptation and association, evaluate the unstructured service interaction values ​​between structured meteorological service business layers. In the formula, This represents the unstructured service interaction value between the e-th language big model node and the f-th language big model node within the x-th response timeframe.

[0049] It should be noted that the structured meteorological service business layer is a binding of "single requirement + single scenario + fixed data channel" (such as "wheat irrigation advice + agricultural scenario + soil moisture channel"), with a fixed structure and single function; however, the actual needs of users are comprehensive (such as farmers need "temperature warning + irrigation advice"), and multiple structured business layers need to cooperate. Since the service types, data dimensions, and scenario logics of different business layers are different (such as "temperature warning" belongs to the disaster warning category, and "irrigation advice" belongs to the agricultural guidance category), there is no "fixed structured rule" for their cooperation relationship (it cannot be defined by a unified table or formula). Therefore, it is necessary to use the "unstructured service interaction value" to quantify this flexible association relationship. The larger the unstructured service interaction value, the higher the overlap degree of the adaptation capabilities of the two nodes within the current time range, and the easier it is to cooperate to provide services. In particular, when the interaction value ≥ the preset threshold (ad), it means that the overlap degree of the adaptation capabilities of the two nodes is high and they can cooperate (such as the high interaction value between "temperature warning" and "irrigation advice", which can be combined into "low temperature + too high soil moisture → cold protection and waterlogging prevention advice"). When the interaction value < ad, it means that there is no overlap in the adaptation boundaries of the two nodes (such as "traffic visibility warning" and "wheat growth monitoring"), and there is no need to cooperate to avoid wasting resources;

[0050] Based on the unstructured service interaction value, it is evaluated whether the structured meteorological service business layers meet the personalized service integration of users, and the personalized service integration degree , where y is the total number of response time ranges, and ad is the preset interaction threshold. If the unstructured service interaction value satisfies , it means that the personalized service integration of users is satisfied, and then is set, otherwise it means that the personalized service integration of users is not satisfied, and then is set;

[0051] It should be noted that the essence of users' personalized needs is "a combination of multiple associated services" (such as travelers need "precipitation + wind force + ultraviolet" services, and farmers need "temperature + soil moisture + crop indicators" services). Since the unstructured service interaction value has quantified the "cooperation feasibility of two nodes", it is possible to judge whether this cooperation "steadily meets the long-term needs of users" by counting the "reach standard situation of interaction values within multiple time ranges";

[0052] Preset the personalized service integration threshold. If the personalized service integration degree is greater than or equal to the personalized service integration threshold, a connection edge is constructed between the e-th language large model node and the f-th language large model node to instruct data sharing between the structured meteorological service business layers, otherwise no connection edge is constructed; based on the result of constructing the connection edge, a meteorological knowledge graph of the language large model nodes is formed as the personalized meteorological service portrait of users;

[0053] It should be noted that "connection edges" are constructed for all nodes that meet the integration standards, and the resulting "node-edge" network is the user's personalized meteorological knowledge graph (such as farmer profiles associated with "temperature warnings, irrigation suggestions, and crop indicators" nodes), realizing "one-stop comprehensive service output".

[0054] Compared with existing technologies, the beneficial effects achieved by this invention are as follows: the construction method sequentially realizes the formation of meteorological data interface sets, the construction and demand parsing of structured meteorological service business layers, the dynamic updating of node service response sets, and the generation of meteorological knowledge graphs and personalized profiles; this invention solves the pain points of existing meteorological service data sources being scattered, demand matching accuracy being low, services lacking timeliness, and lacking personalization by unifying multi-source meteorological data encoding and interface grouping, combining language large models to realize demands, accurately adapting to scenarios, introducing timeliness control to update service matching degree, and constructing a knowledge graph of service node associations, thereby realizing efficient and accurate personalized meteorological services in scenarios such as agriculture, transportation, and tourism. Attached Figure Description

[0055] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0056] Figure 1 This is a schematic diagram illustrating the steps of a method for constructing a meteorological intelligent service based on a large language model according to the present invention. Detailed Implementation

[0057] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. 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 are within the scope of protection of the present invention.

[0058] In this first embodiment: a meteorological intelligent service system based on a large language model is provided. The system includes: a meteorological data integration module, a meteorological service business construction and demand analysis module, a service matching degree recording and response set update module, and a meteorological knowledge graph construction and personalized profile generation module.

[0059] The meteorological data integration module is used to coordinate meteorological data sources, standardized data interfaces, and data transmission channels to form a data interface set.

[0060] The meteorological data integration module includes a data source and interface coordination unit and a data interface set construction unit.

[0061] The data source and interface coordination unit is used to coordinate all meteorological data sources within the meteorological service system, the standardized data interfaces corresponding to each meteorological data source, and the data transmission channels associated with each standardized data interface, and to uniformly encode the meteorological data sources, standardized data interfaces, and data transmission channels respectively;

[0062] The data interface set construction unit is used to group the standardized data interfaces through the data transmission channel to form the data interface set corresponding to the data transmission channel. The data transmission channel includes standardized data interfaces that provide data support for meteorological services.

[0063] In this embodiment, the meteorological data integration module performs unified "type-serial number" encoding on multi-source data (precipitation, temperature, soil moisture, etc.) through the data source and interface coordination unit (e.g., data source D1 = National Meteorological Center, interface I1 = D1-temperature); and standardizes interfaces by data transmission channel by the data interface set construction unit (e.g., channel C1 is associated with the "temperature + precipitation" interface to support agricultural / tourism scenarios), forming a "channel-interface" mapping relationship, providing a structured data foundation for subsequent service calls, and avoiding repeated adaptation of multiple interfaces.

[0064] The meteorological service business construction and requirement parsing module is used to construct the structured meteorological service business layer and parse user service requirements to call the data transmission channel;

[0065] The meteorological service business construction and demand analysis module includes a structured business layer construction unit and a demand analysis and channel call unit.

[0066] The structured business layer construction unit is used to construct a structured meteorological service business layer based on the service requirement parsing capability of the language big model and meteorological service scenarios. One service requirement corresponds to one meteorological service scenario to form a structured meteorological service business layer, and each structured meteorological service business layer is used as a language big model node.

[0067] The requirement parsing and channel invocation unit is used to parse the service requirement parsing capability of the language big model node and adapt it to the meteorological service scenario. It parses the user's service requirements and invokes the corresponding data transmission channel, while capturing the invoked data transmission channel to realize the instruction closed loop of service requirement parsing of the language big model and adaptation to the meteorological service scenario.

[0068] In this embodiment, based on the language big model (training meteorological scenario corpus), the meteorological service business construction and demand parsing module binds "single service demand + single meteorological scenario" to form a structured business layer (i.e., language big model node, such as LLM1 = "wheat greening temperature monitoring + agricultural greening scenario"); at the same time, the demand parsing and channel calling unit captures the called transmission channel to form a closed loop of instructions of "demand parsing → channel calling → data feedback" to ensure accurate matching of demand and data (e.g., parsing the "spring irrigation" demand and automatically calling the "soil moisture + crop index" channel).

[0069] The service matching degree recording and response set update module is used to obtain the service matching degree and update the node service response set;

[0070] The service matching degree recording and response set update module includes a service matching degree acquisition unit and a response set generation and update unit;

[0071] The service matching degree acquisition unit is used to obtain the service matching degree of the data transmission channel in the process of demand analysis and meteorological service scenario adaptation when the language large model node calls the data transmission channel. The service matching degree is the adaptation ratio of the meteorological data source, that is, the ratio of the number of standardized data interfaces corresponding to the meteorological data source adapted to the meteorological service scenario in the data interface set to the total number of standardized data interfaces in the data interface set.

[0072] The response set generation and update unit is used to collect the service matching degree generated when the language big model node calls each data transmission channel and generate the node service response set. At the same time, it pre-constructs the applicable time range segment of the meteorological service scenario, uses this time range segment as the response time range of each structured meteorological service business layer, and updates the node service response set within this response time range.

[0073] In this embodiment, the service matching degree recording and response set update module calculates the service matching degree using the formula "number of interfaces in the data interface set that are adapted to the scenario / total number of interfaces" (e.g., channel C1 has 2 interfaces, and 1 interface is adapted to "temperature requirement", so the matching degree = 0.5), quantifying the degree of data and scenario adaptation. At the same time, it pre-constructs "scenario applicable time range segments" (e.g., typhoon warning time limit of 24 hours, agricultural temperature monitoring time limit of 7 days) as the response time limit, and only updates the node service response set within the time limit to avoid the impact of expired data and ensure service timeliness.

[0074] The meteorological knowledge graph construction and personalized profile generation module is used to quantify the scope of adaptation and association, evaluate the service interaction value and the degree of integration with personalized services, and form a meteorological knowledge graph and a personalized meteorological service profile.

[0075] The meteorological knowledge graph construction and personalized profile generation module includes a quantitative unit for the scope of adaptation and association, a unit for service interaction and integration evaluation, and a knowledge graph and profile generation unit.

[0076] The adaptation association range quantization unit is used to select the maximum and minimum values ​​of service matching degree in the updated node service response set to quantify the upper and lower limits of the adaptation association range of each language large model node.

[0077] The service interaction and integration evaluation unit is used to evaluate the unstructured service interaction value between the structured meteorological service business layers corresponding to the large model nodes of different languages ​​based on the adaptation association range, and evaluate the integration degree of personalized services based on the unstructured service interaction value.

[0078] The knowledge graph and profile generation unit is used to preset the personalized service fusion threshold. If the personalized service fusion degree is greater than or equal to the threshold, a connection edge is built between the corresponding two language big model nodes. Based on the construction results of all connection edges, a meteorological knowledge graph of the language big model node is formed. This meteorological knowledge graph is the user's personalized meteorological service profile.

[0079] In this embodiment, the meteorological knowledge graph construction and personalized profile generation module determines the adaptation boundary of each node (e.g., LLM1 adaptation range = [0, 2.99]) by using the adaptation association range quantification unit and combining the maximum, minimum and statistical values ​​(mean and variance) of the service matching degree. Then, the unstructured service interaction value is calculated by the service interaction and fusion degree evaluation unit (taking the intersection length of the adaptation ranges of two nodes, e.g., the interaction value of LLM1 and LLM2 = 2.23) to quantify the node association strength. The node connection edges are constructed with the condition "personalized service fusion degree ≥ threshold". The resulting meteorological knowledge graph is the user's personalized profile, realizing the association and integration of service nodes (e.g., the farmer profile is associated with the "temperature + precipitation + soil moisture" nodes).

[0080] Please see Figure 1 In this second embodiment: a method for constructing intelligent meteorological services based on a large language model is provided to be applicable to the first embodiment above. In this embodiment, the experimental scenario is "personalized agricultural meteorological services in a major wheat-producing area of ​​a certain city", and the service target is wheat growers.

[0081] The method includes the following steps:

[0082] Step S1: Using structured data, meteorological data sources within the meteorological service system, standardized data interfaces corresponding to each meteorological data source, and data transmission channels associated with each standardized data interface are formed. The standardized data interfaces are then grouped through the data transmission channels to form a data interface set for the data transmission channels.

[0083] For example, all meteorological data sources within the meteorological service system and the standardized data interfaces corresponding to each meteorological data source, as well as the data transmission channels associated with each standardized data interface, are coordinated, and the meteorological data sources, standardized data interfaces, and data transmission channels are uniformly coded respectively;

[0084] The data transmission channel contains standardized data interfaces that provide data support for meteorological services. These standardized data interfaces are grouped together through the data transmission channel to form a set of data interfaces for the data transmission channel.

[0085] For example, when coordinating data sources and interfaces, the data sources include D1 (a city station of the National Meteorological Center), D2 (a city soil monitoring station), and D3 (a city agricultural observation point), uniformly coded as "D + serial number"; the standardized interfaces include I1 (D1 - temperature, JSON format, updated every hour), I2 (D1 - precipitation, JSON, updated every 2 hours), I3 (D2 - soil moisture, JSON, updated every 6 hours), and I4 (D3 - wheat growth adaptation index, XML, updated every 12 hours), coded as "I + serial number"; the data transmission channels include C1 (supporting the "temperature + precipitation" service, associated with I1 and I2) and C2 (supporting the "soil moisture + wheat index" service, associated with I3 and I4), coded as "C + serial number".

[0086] Step S2: Based on the service requirement analysis of the language big model and the meteorological service scenario, construct a structured meteorological service business layer between the service requirements and the meteorological service scenario, which serves as a node of the language big model, and parse the user's service requirements through the language big model node to call the data transmission channel;

[0087] For example, based on the service demand analysis of the language big model and the meteorological service scenario, a structured meteorological service business layer is constructed. In this layer, a service demand and a meteorological service scenario constitute a structured meteorological service business layer, and a structured meteorological service business layer is used as a node of the language big model.

[0088] Based on the service requirement parsing capability and meteorological service scenario adaptation function of the language big model node, each data transmission channel is captured. If the data transmission channel is called after parsing the user's service requirement through the language big model node, the called data transmission channel is captured to realize the instruction closed loop of service requirement parsing of the language big model and meteorological service scenario adaptation.

[0089] For example, construct three large language model nodes: LLM1 requires "low temperature warning during wheat greening" + scenario "wheat greening", associated with channel C1 (requires I1 temperature data); LLM2 requires "rainfall monitoring during wheat grain filling" + scenario "wheat grain filling", associated with channel C1 (requires I2 rainfall data); LLM3 requires "soil moisture recommendations for wheat irrigation" + scenario "wheat irrigation", associated with channel C2 (requires I3 humidity and I4 growth index data).

[0090] When performing requirement parsing and channel invocation, the user inputs the requirement: "I am a wheat farmer in a certain city. I want to know if there will be frost damage during the greening period in the next 7 days, whether there will be rain during the grain filling period, and whether I need to irrigate now?"; Language model parsing: The requirement is broken down into "low temperature warning during greening period (LLM1), precipitation during grain filling period (LLM2), and irrigation suggestions (LLM3)", and channels C1 (supporting LLM1, LLM2) and C2 (supporting LLM3) are automatically invoked; Channel capture: The system captures the invoked C1 and C2 in real time, forming a "parsing-invocation-feedback" closed loop.

[0091] Step S3: Record the service matching degree of the data transmission channel called in the process of demand analysis and meteorological service scenario adaptation to the node service response set, and update the service response set of each node based on the applicable time range of the meteorological service scenario;

[0092] For example, based on the capture behavior of the data transmission channel, when the data transmission channel is called through the language big model node, the service matching degree of the called data transmission channel in the process of demand analysis and meteorological service scenario adaptation is obtained. The service matching degree refers to the adaptation ratio of the meteorological data source. The numerator of the adaptation ratio is the number of standardized data interfaces corresponding to the meteorological data source adapted to the meteorological service scenario in the data interface set, and the denominator of the adaptation ratio is the total number of standardized data interfaces contained in the data interface set.

[0093] Collect the service matching degree generated when the language big model node calls each data transmission channel, and generate the node service response set; pre-build the applicable time range segment of the meteorological service scenario, and use the time range segment as the response time range of each structured meteorological service business layer, and update the service response set of each node within the response time range;

[0094] For example, when calling C1, the interface set S1={I1, I2}: the interface that adapts to LLM1 (temperature requirement) is I1, and the matching degree m(LLM1|C1)=1 / 2=0.5; the interface that adapts to LLM2 (rainfall requirement) is I2, and the matching degree m(LLM2|C1)=1 / 2=0.5; when calling C2, the interface set S2={I3, I4}: the interfaces that adapt to LLM3 (irrigation recommendation) are I3 and I4, and the matching degree m(LLM3|C2)=2 / 2=1.0.

[0095] Step S4: Based on the updated node service response set, quantify the adaptation and association range of language big model nodes, and evaluate the unstructured service interaction value and personalized service integration degree between the structured meteorological service business layers corresponding to different language big model nodes, so as to build connection edges between language big model nodes and form a meteorological knowledge graph of language big model nodes. This meteorological knowledge graph serves as a personalized meteorological service profile for users.

[0096] For example, the maximum and minimum values ​​of service matching degree are selected in the updated service node set to quantify the upper and lower limits of the adaptation association range of the language large model nodes:

[0097] upper limit ;

[0098] lower limit ;

[0099] In the formula, This represents the e-th language large model node. This represents the language large model node within the x-th response time range. The corresponding updated node service response set, and , This represents the r-th data transmission channel. Representation of language large model nodes Invoke data transmission channel The service matching degree generated at that time and These are, in order, the maximum value function and the minimum value function. This represents the mean of the service matching score. For the variance of service matching degree, and , R represents the total number of data transmission channels, and E represents the total number of language large model nodes;

[0100] Then the language large model nodes The adaptation association range within the x-th response timeframe is ;

[0101] Based on the scope of adaptation and association, evaluate the unstructured service interaction values ​​between structured meteorological service business layers. In the formula, This represents the unstructured service interaction value between the e-th language big model node and the f-th language big model node within the x-th response timeframe.

[0102] Based on unstructured service interaction values, assess whether the structured meteorological service business layers meet the requirements for personalized service integration for users, and the degree of personalized service integration. In the formula, y represents the total number of response time ranges, and ad is the preset interaction threshold. If the unstructured service interaction value satisfies... This indicates that the integration of personalized services meets the user's needs, thus enabling... Otherwise, it indicates a failure to meet the user's personalized service integration needs, and thus... ;

[0103] A preset threshold for personalized service integration will be set. If the degree of personalized service integration is... If the value is greater than or equal to the personalized service fusion threshold, a connection edge is constructed between the e-th language big model node and the f-th language big model node to facilitate data sharing between the instruction-structured meteorological service business layers; otherwise, no connection edge is constructed. Based on the result of constructing the connection edge, a meteorological knowledge graph of the language big model nodes is formed as a personalized meteorological service profile for the user.

[0104] For example, the farmer was given a message that said, "The temperature during the greening period in the next 7 days will be 8-15℃, with no risk of low temperature damage; the probability of precipitation during the grain filling period (May 10-20) is 40%, with light to moderate rain around May 15; the current soil moisture is 22% (suitable), and irrigation is not required in the next 3 days."

[0105] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0106] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. 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 intelligent service system based on a large language model, characterized in that, include: The module includes a meteorological data integration module, a meteorological service business construction and demand analysis module, a service matching degree recording and response set update module, and a meteorological knowledge graph construction and personalized profile generation module. The meteorological data integration module is used to coordinate meteorological data sources, standardized data interfaces, and data transmission channels to form a data interface set. The meteorological service business construction and demand parsing module is used to construct a structured meteorological service business layer and parse user service demands to call the data transmission channel; The service matching degree recording and response set update module is used to obtain the service matching degree and update the node service response set; The meteorological knowledge graph construction and personalized profile generation module is used to quantify the scope of adaptation and association, evaluate the service interaction value and the degree of integration of personalized services, and form a meteorological knowledge graph and a personalized meteorological service profile. The meteorological data integration module includes a data source and interface coordination unit and a data interface set construction unit; The data source and interface coordination unit is used to coordinate all meteorological data sources within the meteorological service system, the standardized data interfaces corresponding to each meteorological data source, and the data transmission channels associated with each standardized data interface, and to uniformly encode the meteorological data sources, standardized data interfaces, and data transmission channels respectively. The data interface set construction unit is used to group each standardized data interface through the data transmission channel to form a data interface set corresponding to the data transmission channel, wherein the data transmission channel includes standardized data interfaces that provide data support for meteorological services. The meteorological service business construction and demand analysis module includes a structured business layer construction unit and a demand analysis and channel invocation unit; The structured business layer construction unit is used to construct a structured meteorological service business layer based on the service requirement parsing capability of the language big model and the meteorological service scenario. One service requirement corresponds to one meteorological service scenario to form a structured meteorological service business layer, and each structured meteorological service business layer is used as a language big model node. The requirement parsing and channel invocation unit is used to parse the user's service requirements and invoke the corresponding data transmission channel based on the service requirement parsing capability of the language big model node and the weather service scenario adaptation function. At the same time, it captures the invoked data transmission channel to realize the instruction closed loop of service requirement parsing of the language big model and weather service scenario adaptation. The service matching degree recording and response set update module includes a service matching degree acquisition unit and a response set generation and update unit; The service matching degree acquisition unit is used to acquire the service matching degree of the data transmission channel in the process of demand analysis and meteorological service scenario adaptation when the language large model node calls the data transmission channel. The service matching degree is the adaptation ratio of the meteorological data source, that is, the ratio of the number of standardized data interfaces corresponding to the meteorological data source adapted to the meteorological service scenario in the data interface set to the total number of standardized data interfaces in the data interface set. The response set generation and update unit is used to collect the service matching degree generated when the language big model node calls each data transmission channel and generate the node service response set. At the same time, it pre-constructs the applicable time range segment of the meteorological service scenario, uses the time range segment as the response time range of each structured meteorological service business layer, and updates the node service response set within the response time range. The meteorological knowledge graph construction and personalized profile generation module includes an adaptation and association range quantification unit, a service interaction and integration degree evaluation unit, and a knowledge graph and profile generation unit. The adaptation association range quantization unit is used to select the maximum and minimum values ​​of service matching degree in the updated node service response set, respectively, so as to quantify the upper and lower limits of the adaptation association range of each language large model node. The service interaction and integration evaluation unit is used to evaluate the unstructured service interaction value between the structured meteorological service business layers corresponding to different language large model nodes based on the adaptation association range, and evaluate the personalized service integration degree based on the unstructured service interaction value. The knowledge graph and profile generation unit is used to preset a personalized service fusion threshold. If the personalized service fusion degree is greater than or equal to the threshold, a connection edge is constructed between the corresponding two language large model nodes. Based on the construction results of all connection edges, a meteorological knowledge graph of the language large model nodes is formed. This meteorological knowledge graph is the user's personalized meteorological service profile.

2. A method for constructing intelligent meteorological services based on a large language model, applied to the intelligent meteorological service system based on a large language model as described in claim 1, characterized in that, include: Step S1: Using structured data, meteorological data sources within the meteorological service system, standardized data interfaces corresponding to each meteorological data source, and data transmission channels associated with each standardized data interface are formed. The standardized data interfaces are then grouped through the data transmission channels to form a data interface set for the data transmission channels. Step S2: Based on the service requirement analysis of the language big model and the meteorological service scenario, construct a structured meteorological service business layer between the service requirements and the meteorological service scenario, which serves as a node of the language big model, and parse the user's service requirements through the language big model node to call the data transmission channel; Step S3: Record the service matching degree of the data transmission channel called in the process of demand analysis and meteorological service scenario adaptation to the node service response set, and update the service response set of each node based on the applicable time range of the meteorological service scenario; Step S4: Based on the updated node service response set, quantify the adaptation and association range of the language big model nodes, and evaluate the unstructured service interaction value and personalized service integration degree between the structured meteorological service business layers corresponding to different language big model nodes, so as to build connection edges between language big model nodes and form a meteorological knowledge graph of language big model nodes. This meteorological knowledge graph serves as a personalized meteorological service profile for users.

3. The method for constructing a meteorological intelligent service based on a large language model according to claim 2, characterized in that, The specific implementation process of step S1 includes: The system coordinates all meteorological data sources within the meteorological service system, as well as the standardized data interfaces corresponding to each meteorological data source and the data transmission channels associated with each standardized data interface, and uniformly encodes the meteorological data sources, standardized data interfaces, and data transmission channels respectively; The data transmission channel includes standardized data interfaces that provide data support for meteorological services. These standardized data interfaces are grouped together through the data transmission channel to form a data interface set for the data transmission channel.

4. The method for constructing a meteorological intelligent service based on a large language model according to claim 2, characterized in that, The specific implementation process of step S2 includes: Based on the service demand analysis of the language big model and the meteorological service scenario, a structured meteorological service business layer is constructed. In this layer, a service demand and a meteorological service scenario constitute a structured meteorological service business layer, and a structured meteorological service business layer is used as a node of the language big model. Based on the service requirement parsing capability and meteorological service scenario adaptation function of the language big model node, each data transmission channel is captured. If a data transmission channel is called after parsing the user's service requirement through the language big model node, the called data transmission channel is captured to realize the instruction closed loop of service requirement parsing of the language big model and meteorological service scenario adaptation.

5. The method for constructing a meteorological intelligent service based on a large language model according to claim 2, characterized in that, The specific implementation process of step S3 includes: Based on the capture behavior of the data transmission channel, when the data transmission channel is called through the language large model node, the service matching degree of the called data transmission channel in the process of demand analysis and meteorological service scenario adaptation is obtained. The service matching degree refers to the adaptation ratio of the meteorological data source. The numerator of the adaptation ratio is the number of standardized data interfaces corresponding to the meteorological data source adapted to the meteorological service scenario in the data interface set, and the denominator of the adaptation ratio is the total number of standardized data interfaces included in the data interface set. Collect the service matching degree generated when the language large model node calls each data transmission channel, and generate a node service response set; pre-construct the applicable time range segment of the meteorological service scenario, and use the time range segment as the response time range of each structured meteorological service business layer, and update the service response set of each node within the response time range.

6. The method for constructing a meteorological intelligent service based on a large language model according to claim 2, characterized in that, The specific implementation process of step S4 includes: In the updated service node set, the maximum and minimum service matching scores are selected respectively to quantify the upper and lower limits of the adaptation and association range of the language large model nodes: ； In the formula, This represents the e-th language large model node. This represents the language large model node within the x-th response time range. The corresponding updated node service response set, and , This represents the r-th data transmission channel. Representation of language large model nodes Invoke data transmission channel The service matching degree generated at that time and These are, in order, the maximum value function and the minimum value function. This represents the mean of the service matching score. For the variance of service matching degree, and , R represents the total number of data transmission channels, and E represents the total number of language large model nodes; Then the language large model nodes The adaptation association range within the x-th response timeframe is ; Based on the scope of adaptation and association, evaluate the unstructured service interaction values ​​between structured meteorological service business layers. In the formula, This represents the unstructured service interaction value between the e-th language big model node and the f-th language big model node within the x-th response timeframe. Based on unstructured service interaction values, assess whether the structured meteorological service business layers meet the requirements for personalized service integration for users, and the degree of personalized service integration. In the formula, y represents the total number of response time ranges, and ad is the preset interaction threshold. If the unstructured service interaction value satisfies... This indicates that the integration of personalized services meets the user's needs, thus enabling... Otherwise, it indicates a failure to meet the user's personalized service integration needs, and thus... ; A preset threshold for personalized service integration will be set. If the degree of personalized service integration is... If the value is greater than or equal to the personalized service integration threshold, a connection edge is constructed between the e-th language big model node and the f-th language big model node to facilitate data sharing between the instruction-structured meteorological service business layers; otherwise, no connection edge is constructed. Based on the result of constructing the connection edge, a meteorological knowledge graph of the language big model nodes is formed as a personalized meteorological service profile for the user.

Images