Large model assisted decision-making method and system applied to natural resource informatization management

Abstract

The application provides a large model assisted decision-making method and system applied to natural resource informatization management, relates to the technical field of natural resource informatization management, first receives and identifies the core demand input of a natural resource management scene, obtains type characteristic data and real-time change data flow of the natural resource, generates a scene demand analysis result through a pre-trained natural language processing model and a knowledge graph, generates a large model decision-making adaptation rule through a rule construction engine in combination with a basic specification document library and a historical decision-making case database, drives a natural resource management large model to output a management action combination scheme by using the large model decision-making adaptation rule, issues the management action combination scheme to an execution unit, collects multidimensional data in an execution process to generate a feedback data set, and finally iteratively optimizes the large model decision-making adaptation rule and the management action combination scheme by using the feedback data set. The application improves the efficiency and effect of natural resource informatization management decision-making.

Description

Technical Field

[0001] This invention relates to the field of natural resource information management technology, and more specifically, to a large-scale model-assisted decision-making method and system for natural resource information management. Background Technology

[0002] In the field of information management of natural resources, with the increasing variety of natural resources, the growing complexity of management scenarios, and the continuous changes in management needs, traditional management decision-making methods are facing many challenges.

[0003] Currently, natural resource management decisions mainly rely on human experience and some relatively simple data analysis tools. While human experience can address common management problems to a certain extent, its limitations are becoming increasingly apparent in the face of complex and ever-changing natural environments, policy requirements, and technological advancements. For example, human decision-making struggles to quickly process massive amounts of natural resource data and cannot capture real-time dynamic changes in natural resources, thus affecting the timeliness and accuracy of decisions.

[0004] Meanwhile, existing data analysis tools often only process specific data types and simple management needs, lacking a comprehensive understanding and in-depth analysis of the core requirements of natural resource management scenarios. They struggle to comprehensively consider multiple factors, including the characteristics of natural resource types, real-time data streams, and policy requirements, to generate scientifically sound management decision-making solutions. Furthermore, during the implementation of these solutions, there is a lack of effective feedback mechanisms to evaluate and optimize the decision-making process, making it difficult to adapt to the ever-changing needs of natural resource management. Summary of the Invention

[0005] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide a large-scale model-assisted decision-making method for natural resource information management, the method comprising:

[0006] The system receives and identifies the core requirements input of a natural resource management scenario, and acquires the type feature data and real-time change data stream of natural resources. The core requirements input, type feature data, and real-time change data stream are then input into a pre-trained natural language processing model and knowledge graph for entity recognition, relation extraction, and constraint extraction to generate scenario requirement analysis results. The scenario requirement analysis results include a management target direction data list and a scenario constraint data list. The core requirements input includes a target description text of natural resource management, a set of scenario-related natural condition data, policy requirement documents, and a set of technical application condition parameters.

[0007] Based on the scenario requirement analysis results, a large model decision adaptation rule is generated. The scenario requirement analysis results, the basic normative document library of natural resource management, and the historical decision case database are taken as input and processed by the rule building engine to generate the large model decision adaptation rule. The large model decision adaptation rule includes a data table of the mapping relationship between requirements and decisions and a list of execution constraints.

[0008] The large model decision adaptation rules drive the pre-trained natural resource management large model to output management action combination schemes. The large model decision adaptation rules and the current status data stream of natural resources obtained from the monitoring terminal are input into the decision generation module of the large model for processing to generate management action combination schemes. The management action combination schemes include specific management operation instruction sequences and execution timing schedule tables.

[0009] The management action combination scheme is issued to the execution unit in the natural resource management scenario. The data stream of resource status change, operation execution effect log and scenario environment change information during the execution process are collected by the sensor network and execution log collection system deployed in the scenario. After cleaning, fusion and formatting, a multi-dimensional feedback dataset is generated. The multi-dimensional feedback dataset includes a set of execution effect indicators and a set of scenario change features.

[0010] The large model decision adaptation rules and the management action combination scheme are iteratively optimized using the multi-dimensional feedback dataset. The multi-dimensional feedback dataset, the historical versions of the large model decision adaptation rules, and the historical versions of the management action combination scheme are input into the iterative optimization algorithm module for processing, and an optimized decision scheme data package is generated.

[0011] Furthermore, embodiments of the present invention also provide a large-scale model-assisted decision-making system for natural resource information management, comprising:

[0012] A processor; a machine-readable storage medium for storing machine-executable instructions of the processor; wherein the processor is configured to execute the aforementioned large-scale model-assisted decision-making method for natural resource information management by executing the machine-executable instructions.

[0013] In another aspect, embodiments of the present invention also provide a computer program product, the computer program product including machine-executable instructions, the machine-executable instructions being stored in a computer-readable storage medium, a processor of a large model-assisted decision-making system for natural resource information management reading the machine-executable instructions from the computer-readable storage medium, the processor executing the machine-executable instructions, causing the large model-assisted decision-making system for natural resource information management to execute the aforementioned large model-assisted decision-making method for natural resource information management.

[0014] Based on the above, by comprehensively and accurately receiving and identifying the core demand inputs of natural resource management scenarios, and simultaneously acquiring type characteristic data and real-time change data streams of natural resources, a pre-trained natural language processing model and knowledge graph are used to deeply analyze the core demand inputs, type characteristic data, and real-time change data streams. This generates scenario demand analysis results containing a data list of management objective directions and a data list of scenario constraints. Based on the scenario demand analysis results, a large-scale model decision adaptation rule is generated. Combined with a basic normative document library and historical decision case database for natural resource management, a rule building engine generates a data table containing the mapping relationship between demands and decisions and a list of execution constraints, which can better adapt to the needs of different natural resource management scenarios. The large-scale model decision adaptation rule drives the pre-trained natural resource management large-scale model to output a management action combination scheme. This scheme includes a specific management operation instruction sequence and an execution timing schedule table, providing detailed and feasible operational guidance for natural resource management and improving the execution efficiency of management decisions. After the management action combination plan is issued to the execution unit, multi-dimensional data during the execution process is collected through the sensor network and execution log collection system, and a multi-dimensional feedback dataset is generated. This feedback dataset is used to iteratively optimize the large model decision adaptation rules and management action combination plan. The decision plan can be continuously adjusted according to the actual execution effect, making the decision more in line with the actual situation of natural resource management, and effectively improving the decision-making level and management effect of natural resource information management. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the execution flow of the large-scale model-assisted decision-making method for natural resource information management provided in this embodiment of the invention.

[0016] Figure 2 This is a schematic diagram of exemplary hardware and software components of a large-scale model-assisted decision-making system for information management of natural resources, provided in an embodiment of the present invention. Detailed Implementation

[0017] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 This is a flowchart illustrating a large-scale model-assisted decision-making method for natural resource information management, provided by an embodiment of the present invention. The following is a detailed description of this large-scale model-assisted decision-making method for natural resource information management.

[0018] Step S110: Receive and identify the core requirement input of the natural resource management scenario, and obtain the type feature data and real-time change data stream of natural resources; input the core requirement input, type feature data, and real-time change data stream into a pre-trained natural language processing model and knowledge graph for entity recognition, relation extraction, and constraint extraction to generate scenario requirement analysis results. The scenario requirement analysis results include a management target direction data list and a scenario constraint data list. The core requirement input includes the target description text of natural resource management, the scenario-related natural condition data set, policy requirement documents, and technology application condition parameter set.

[0019] In this embodiment, the sustainable management of forest resources in a provincial-level forest area is taken as the application scenario. First, core requirements input is received via a network interface, which includes target description text, a set of natural condition data, policy requirement documents, and a set of technical application condition parameters. The target description text is a natural language description of forest resource management; the natural condition data set covers information on the forest area's topography, climate, and soil; the policy requirement documents contain normative documents related to forest resource management; and the technical application condition parameter set involves the performance indicators of existing technical equipment. Simultaneously, type characteristic data of natural resources is obtained from databases and monitoring equipment. This type characteristic data includes information such as the composition of the main tree species in the forest area, the growth status of each tree species, and their distribution range. The real-time change data stream includes real-time forest area environmental data collected by sensors, such as temperature, humidity, and light intensity, as well as real-time monitoring data on the occurrence of pests and diseases.

[0020] The core requirements, type feature data, and real-time change data streams obtained above are input into a pre-trained natural language processing (NLP) model and knowledge graph. The NLP model adopts a Transformer architecture, containing multiple encoder layers, each composed of a multi-head self-attention mechanism and a feedforward neural network. When processing text data, word segmentation is performed first, dividing the text into multiple words or sub-word units, and then generating corresponding word vectors for each word or sub-word unit. The encoder layers process the word vector sequences to capture the contextual semantic information in the text. The entity recognition module identifies key entities in the text based on the processed text features, such as "forest coverage," "biodiversity," and "pine wilt disease." The relation extraction module analyzes the semantic relationships between the identified entities, such as "increase - forest coverage" and "prevent - pine wilt disease." The constraint extraction module extracts conditions that restrict forest resource management decisions from policy requirement documents, such as "logging operations are prohibited in specific areas" and "logging volume must not exceed a certain limit." The knowledge graph contains various entities in the forest resource domain and their relationships, used to assist in entity identification and relation extraction, improving the accuracy of parsing. By associating the output of the natural language processing model with the knowledge graph, the entity and relationship information is further enriched and improved, ultimately generating the scenario requirement parsing result. The management objective direction data list in this scenario requirement parsing result clarifies the main objectives of forest resource management, such as increasing forest coverage and protecting biodiversity; the scenario constraint data list lists the various constraints that need to be followed in achieving these objectives.

[0021] Step S120: Generate large model decision adaptation rules based on the scenario requirement analysis results. The scenario requirement analysis results, the basic normative document library of natural resource management, and the historical decision case database are taken as input and processed by the rule building engine to generate large model decision adaptation rules. The large model decision adaptation rules include a data table of the mapping relationship between requirements and decisions and a list of execution constraints.

[0022] The rule-building engine receives scenario requirement analysis results, a basic normative document library for natural resource management, and a historical decision-making case database as input. The basic normative document library stores various norms and standards related to forest resource management, while the historical decision-making case database contains past decision-making cases made in similar scenarios and their implementation effects. The rule-building engine processes this input data. First, it analyzes the scenario requirement analysis results to clarify management objectives and constraints. Then, it matches the scenario requirement analysis results with the normative clauses in the basic normative document library to identify normative content relevant to the current requirement. Simultaneously, it retrieves cases similar to the current scenario from the historical decision-making case database and analyzes the decision-making processes and basis in these cases. By comprehensively considering scenario requirements, basic norms, and historical cases, the rule-building engine generates large-scale model decision adaptation rules. These rules include a mapping table between requirements and decisions, illustrating which decision-making schemes correspond to different management requirements; and a list of execution constraints, specifying the conditions that must be met during the execution of the decision-making scheme.

[0023] Step S121: Perform semantic segmentation and clustering operations on the management target direction data list in the scenario requirement analysis result, split it into multiple specific target item data, and attach a performance-oriented label predefined by the system to each specific target item data. At the same time, extract the scenario constraint data list from the scenario requirement analysis result, and label each constraint data with an applicable scenario scope label predefined by the system and an impact dimension label predefined by the system.

[0024] The management objective direction data list from the scenario requirement analysis results is processed first through semantic segmentation. Natural language processing (NLP) techniques are used to analyze the descriptive text of each management objective direction, breaking it down into multiple specific and actionable target items. For example, the management objective direction of "increasing forest coverage" is broken down into specific target items such as "increasing afforestation area," "improving tree survival rate," and "reducing deforestation." Then, clustering is performed. Based on the semantic similarity and relevance of the specific target item data, similar specific target item data are grouped together to better organize and manage the objectives.

[0025] Achievement-oriented labels are attached to each specific target data item. These labels are predefined based on the nature of the specific target item and the expected outcome, such as "growth-oriented," "stabilizing," or "optimizing." By analyzing the textual descriptions of the specific target data, the type of achievement-oriented label it belongs to is determined, enabling the selection of appropriate decision-making strategies based on the label type during subsequent decision-making processes.

[0026] Simultaneously, a list of scenario constraint data is extracted from the scenario requirement analysis results. For each constraint data, its descriptive text is analyzed to determine its applicable scenario scope, and corresponding applicable scenario scope labels are added, such as "specific region" or "specific time period." Furthermore, the impact of the constraints on decision-making is analyzed, and impact dimension labels are added, such as "resource quantity," "environmental factors," and "technical conditions."

[0027] Step S1211: Extract the scenario requirement information field of each record in the historical decision case database, and parse the management objective description text and scenario restriction description text of each historical decision case from the scenario requirement information field.

[0028] Access the historical decision-making case database to retrieve each historical decision-making case record. For each record, locate the scenario requirement information field, which stores information related to the scenario requirements faced during the decision-making process. Parse the content of the scenario requirement information field, using natural language processing technology to identify and extract the management objective description text and scenario constraint description text. The management objective description text reflects the decision-making objective of the historical case, while the scenario constraint description text reflects the constraints encountered during the decision-making and execution process.

[0029] Step S1212: Calculate the semantic similarity between the management objective description text of each historical decision case and the management objective direction data list in the scenario requirement analysis result. According to the preset similarity threshold, the historical decision cases are initially divided into core objective matching case set, partial objective matching case set and objective mismatch case set, and the objective mismatch case set is removed from the current processing queue.

[0030] Semantic similarity is calculated between the management objective description text of each historical decision case and the management objective direction data list in the scenario requirement analysis results. The semantic similarity calculation uses a pre-trained language model-based method to convert the text into vector representations and then calculates the cosine similarity between the vectors. Based on the calculated semantic similarity values ​​and a preset similarity threshold, historical decision cases are classified. When the similarity value is higher than a higher threshold, the case is classified as a core objective matching case set; when the similarity value is in a medium range, it is classified as a partially objective matching case set; and when the similarity value is lower than a lower threshold, it is classified as an objective mismatch case set. Objective mismatch case sets, due to their significant difference from the current scenario requirements, are removed from the current processing queue and no longer participate in subsequent analysis and processing.

[0031] Step S1213: Extract key restriction feature words from the scenario restriction description text of each case in the initially divided core target matching case set and partial target matching case set, and perform feature matching calculation with the scenario restriction condition data list in the scenario requirement analysis result.

[0032] For each case in the core target matching case set and the partial target matching case set, key restriction feature words are extracted from its scenario restriction description text. The extraction of key restriction feature words employs text analysis techniques, including word segmentation, stop word removal, and keyword extraction, to obtain feature words that reflect the core content of the restriction conditions. The extracted key restriction feature words are then matched with features in the scenario restriction data list from the scenario requirement analysis results. Matching calculations can be performed by calculating feature word overlap and semantic similarity to determine the degree of matching between the scenario restrictions of historical cases and the current scenario requirement restrictions.

[0033] Step S1214: Based on the matching calculation results of key restriction feature words, the core target matching case set is further divided into a fully restricted matching case subset and a partially restricted matching case subset. The partially target matching case set is further divided into a partially target matching case subset where the number of key restriction feature matches reaches a preset threshold and a partially target matching case subset where the number of key restriction feature matches does not reach the preset threshold.

[0034] Based on the matching results of key restrictive feature words, the core target matching case set and the partial target matching case set are further divided. For the core target matching case set, when the matching degree of key restrictive feature words reaches a high level, such as when the matching degree exceeds a preset high threshold, it is divided into a fully restricted matching case subset; when the matching degree is at a medium level, that is, higher than a preset low threshold but lower than a preset high threshold, it is divided into a partially restricted matching case subset.

[0035] For a subset of target matching cases, the number of key constraint feature matches for each case is counted. Cases with a key constraint feature match count reaching a preset threshold are classified as a subset of target matching cases with a key constraint feature match count reaching the preset threshold; while cases with a match count not reaching the preset threshold are classified as a subset of target matching cases with a key constraint feature match count not reaching the preset threshold.

[0036] Step S1215: Extract the data features of each type of case subset after division, record the common decision-making logic pattern, typical execution path sequence, common operation type set and core performance indicator data of each case subset, and form a category feature description file.

[0037] Data features are extracted from the categorized case subsets. Historical decision-making cases within each subset are analyzed to summarize the common decision-making logic patterns they follow, i.e., the general logic and thinking methods used in making decisions. Typical execution path sequences are identified, i.e., the steps and process order typically followed by the cases when implementing decision-making plans. Common operation types are summarized, i.e., the specific operational methods and approaches frequently used in the cases. Core performance indicators are extracted, reflecting the effects of the decisions implemented, such as the rate of change in forest cover and the degree of improvement in tree growth. These extracted data features are organized and recorded to form a category feature description file.

[0038] Step S1216: Based on the category feature description file, assign a system-generated category identifier and a feature-generated applicable scenario description text to each subset of historical decision cases.

[0039] Based on the descriptions of the characteristics of various historical decision case subsets in the category feature description file, a unique category identifier is generated for each case subset to distinguish and identify different case subsets in the system. Simultaneously, an applicable scenario description text is generated based on the characteristics of the case subset. This scenario requirement information text details the applicable scenario conditions for the case subset, including characteristics such as management objectives, constraints, and environmental factors. By assigning category identifiers and generating applicable scenario description texts, case subsets can be managed and queried more conveniently, and it is possible to quickly determine whether a case subset is suitable for the current scenario requirements.

[0040] Step S1217: Sort historical decision cases under the same category by time according to the decision timestamp, analyze the adjustment trend of decision parameters of historical decision cases in different time periods, and summarize the data on the influence of time factors on decision parameters.

[0041] For historical decision-making cases within the same category, they are sorted chronologically based on the decision timestamps in the case records, arranging the cases in chronological order of decision time. Then, the decision parameters for these cases at different time periods are analyzed to observe how these parameters change over time, summarizing the adjustment trends of the decision parameters. For example, decision parameters such as afforestation area and logging volume may have different values ​​in different seasons or years. By analyzing these trends, the influence pattern data of time factors on decision parameters is summarized, revealing how time factors affect the selection and adjustment of decision parameters.

[0042] Step S1218: Extract typical historical decision cases whose decision performance index data are higher than a preset threshold from each subset of historical decision cases, and record the unique identifier and key decision steps sequence of the typical historical decision cases.

[0043] Within each subset of historical decision-making cases, preset thresholds are set for decision performance indicators. By comparing the performance indicator data of the cases with these preset thresholds, cases with indicator data exceeding the preset thresholds are selected. These cases are considered typical historical decision-making cases with good execution results. Unique identifiers for these typical cases are recorded to facilitate quick location and retrieval of specific case information. Simultaneously, the key decision-making step sequences of typical historical decision-making cases are extracted—that is, the steps and process sequences that play a crucial role in the decision-making process. These key step sequences reflect the critical links in successful decision-making and have significant reference value for generating effective decision-making solutions in the current scenario.

[0044] Step S1219: Perform reasoning and completion operations based on association rules on the missing key information fields in the records of each category of historical decision-making case subsets. Based on the completed decision-making process data, execution detail data, and effect feedback data of each category of historical decision-making case records, integrate them according to the classification labels to form a case classification index.

[0045] Examine the key information fields of the records in each category of historical decision-making case subsets to identify missing fields. Employ a reasoning-based completion method based on association rules to analyze the relationships between other relevant fields in the case subsets and establish association rules. Using these association rules, infer reasonable values ​​for the missing fields based on existing field information, thus completing the missing key information fields. The completed historical decision-making case records contain complete decision-making process data, execution details data, and effect feedback data. Integrate the completed case records according to the case category tags to construct a case category index. The case category index allows users to quickly find and retrieve relevant historical decision-making cases by category tags, improving case utilization efficiency.

[0046] Step S122: Retrieve the basic normative document library for natural resource management. The basic normative document library for natural resource management includes a collection of documents consisting of resource protection norms, development and utilization norms, ecological maintenance norms, and supervision and management norms. Use a text matching algorithm to scan each clause in the basic normative document library, calculate the correlation score between each clause and each data item in the current scenario requirement analysis result, and filter out normative clauses with a correlation score higher than a preset threshold.

[0047] The system database retrieves a basic normative document library for natural resource management. This library contains various types of normative documents, such as those for resource protection, development and utilization, ecological maintenance, and supervision and management. A text matching algorithm is used to scan and analyze each normative clause in the library. Each clause is compared with data from the current scenario requirement analysis results, and a relevance score is calculated. The relevance score is based on factors such as semantic similarity and keyword matching between the clause and the scenario requirement data. A preset relevance threshold is set, and clauses with relevance scores higher than this threshold are selected.

[0048] Step S123: Obtain the historical decision case database, perform cluster analysis on the records in the historical decision case database, and automatically classify and index the historical decision case records according to the label system of the management target direction data list and the scenario constraint condition data list. The historical decision case database contains historical decision instance records with similar characteristics to the current management scenario. Each historical decision case record contains scenario requirement information field, decision content field, execution process log field, and final effect indicator field.

[0049] A historical decision-making case database was acquired, containing records of historical decisions that share similar characteristics with the current natural resource management scenario. Each case record includes detailed information such as scenario requirements, decision content, execution process logs, and final outcome metrics. Cluster analysis was performed on these historical decision-making case records, using a label system based on a list of management objective directions and a list of scenario constraints as the classification criteria. Clustering algorithms grouped case records with similar management objectives and scenario constraints into one category. An index was built on the categorized case records to enable quick and accurate retrieval of historical cases relevant to the current scenario requirements. The index was constructed based on the case's classification labels and key features, improving the efficiency and accuracy of case retrieval.

[0050] Step S124: Associate and map the split specific target item data with the filtered normative clauses. Based on the performance-oriented labels of the specific target item data, bind them with the normative clauses that meet their constraints. Establish the mapping relationship between each specific target item data and at least one normative clause through traversal mapping operations, and generate a mapping relationship table.

[0051] The specific target item data obtained in step S121 is associated and mapped with the normative clauses selected in step S122. Based on the outcome-oriented labels of the specific target item data, the nature and requirements of each specific target item are analyzed to identify the normative clauses that can satisfy the constraints of that target. The specific target item data is bound to the corresponding normative clauses to ensure that the achievement of the target complies with the requirements of the relevant norms. By traversing all specific target item data and normative clauses, a mapping relationship is established between them one by one to ensure that each specific target item data is associated with at least one normative clause. The above mapping relationships are compiled into a mapping relationship table, which shows the correspondence between specific target items and normative clauses.

[0052] Step S125: Perform pattern mining on the classified historical decision-making case records, extract the decision logic graph and execution critical path of each type of historical decision-making case, analyze the association pattern between the scenario requirement information field and the decision content field in the historical decision-making case, summarize the adjustment pattern of the decision content field under different scenario constraints, and store the summarized patterns as a historical decision-making case decision pattern knowledge base.

[0053] Pattern mining is performed on categorized historical decision-making case records. Data mining techniques are used to deeply analyze the decision-making process and execution in various case records. Decision logic graphs for each type of historical decision-making case are extracted, illustrating the logical relationships and mutual influences between various factors in the decision-making process. Critical execution paths are identified, i.e., the sequence of steps that plays a crucial role in the final outcome when implementing the decision plan. The association patterns between scenario requirement information fields and decision content fields in historical decision-making cases are analyzed to understand the formation methods and basis of decision content under different scenario requirements. Adjustment patterns of decision content fields under different scenario constraints are summarized, i.e., how decision content is adjusted accordingly to adapt to new conditions when scenario constraints change. These summarized decision patterns are stored to construct a historical decision-making case decision pattern knowledge base.

[0054] Step S126: Based on the mapping relationship table and the historical decision case decision pattern knowledge base, an initial decision rule set is constructed by filling and combining rule templates. The initial decision rule set includes rule structure element data, the relationship data of each element, and the definition data of the scope of application of the rules.

[0055] Combining information from the mapping relationship table and the historical decision-making case decision-making pattern knowledge base, an initial set of decision-making rules is constructed using rule templates. The rule templates contain the basic structure and elements of the decision-making rules, such as condition sections, action sections, and conclusion sections. Based on the mapping relationship between specific target items and normative clauses in the mapping relationship table, and the decision-making patterns in the historical decision-making case decision-making pattern knowledge base, relevant information is populated into the rule templates. The populated rule templates are then combined and adjusted to ensure logical consistency and completeness among the rules. The constructed initial set of decision-making rules includes rule structure element data, i.e., the various components of the rules; correlation data of the elements, i.e., the interrelationships and influences between rule elements; and rule scope definition data, i.e., the scenarios, conditions, and scope to which the rules apply.

[0056] For example, step S1261: Analyze the specific target item data and corresponding specification clauses in the mapping relationship table, and extract the core decision element features of each specific target item data. The core decision element features include target-oriented labels, specification requirement text, execution boundary conditions, data support parameters and key content related to decision.

[0057] A detailed analysis was conducted on the specific target item data and corresponding regulatory clauses in the mapping table. For each specific target item data, its core decision-making element characteristics were extracted. These core decision-making element characteristics include: target-oriented labels, which reflect the nature of the specific target item and the expected effect; regulatory requirement text, i.e., the specific requirements and provisions for the target in the regulatory clauses; execution boundary conditions, which clarify the limitations and conditions in the target execution process; data supporting parameters, i.e., the data indicators and parameters that support the decision; and other key content related to the decision.

[0058] Step S1262: Extract the decision logic graphs and critical path sequences of various historical decision cases from the historical decision case decision pattern knowledge base, summarize the decision step sequences and operation combination patterns under different goal orientation labels, and form a decision process pattern library.

[0059] Access the historical decision-making case study knowledge base to extract decision logic diagrams and critical path sequences for various historical decision-making cases. Classify and organize these decision logic diagrams and critical path sequences based on goal-oriented tags. Summarize common decision-making step sequences under different goal-oriented tags, i.e., the order of decision-making steps typically followed to achieve a specific goal; and operational combination patterns, i.e., the combination methods and synergistic relationships between different operations.

[0060] Step S1263: Combining the characteristics of core decision elements and the decision process pattern library, determine the core logical structure of the initial decision rule set, including the hierarchical structure definition of the rules, the core functional functions of each level, and the data flow interface definition between levels.

[0061] By combining the extracted core decision-making element features with the content in the decision-making process model library, and comprehensively considering factors such as decision-making objectives, normative requirements, decision-making steps, and operational modes, the core logical structure of the initial decision-making rule set is determined. The core logical structure includes the hierarchical definition of the rules, dividing the decision-making rules into different levels, such as the objective layer, strategy layer, and operational layer, with each level responsible for different decision-making functions. The core functional functions of each level are clearly defined, that is, the specific methods and algorithms for implementing the core decision-making functions in each level. Simultaneously, data flow interfaces between levels are defined, specifying the methods, formats, and content of data transmission between different levels to ensure smooth data flow and interaction within the rule set, guaranteeing the effective execution of the decision-making rules.

[0062] Step S1264: Use the specific target item data as the top-level trigger condition predicate of the initial decision rule set, with each specific target item data corresponding to a rule branch path.

[0063] The specific target item data, after being broken down, serves as the top-level trigger condition predicate for the initial decision rule set. When the system receives a specific management request, it triggers the corresponding rule branch path based on the specific target item data. Each specific target item data corresponds to an independent rule branch path, which contains a series of decision rules and operational steps formulated to achieve that specific target. By using the specific target item data as the top-level trigger condition, it ensures that the decision rule set can accurately respond to and process different target needs, improving the targeting and efficiency of decision-making.

[0064] Step S1265: Fill the content of the specification clauses in the mapping relationship table into the corresponding rule branch paths, and record the execution constraint data in each rule branch path.

[0065] Based on the correspondence between specific target items and regulatory clauses in the mapping table, the content of the regulatory clauses is populated into the corresponding rule branch paths. In each rule branch path, execution constraint data is recorded. This data, derived from the regulatory clauses, clarifies the restrictions and requirements that must be followed when executing decisions and operations within that rule branch path. By populating the regulatory clause content and recording execution constraint data, it is ensured that the decisions and operations within the rule branch paths comply with relevant regulations, guaranteeing the legality and compliance of the decisions.

[0066] Step S1266: Referring to the decision logic graph in the historical decision case decision pattern knowledge base, generate a decision reasoning chain data structure for each rule branch path, which contains several sequentially executed judgment nodes and execution nodes, and record the key judgment node functions and required data input interfaces in the decision reasoning chain.

[0067] Referring to the decision logic graph in the historical decision-making case knowledge base, a decision reasoning chain data structure is constructed for each rule branch path. The decision reasoning chain consists of several sequentially executed judgment nodes and execution nodes. Judgment nodes are used to make judgments and decisions based on input data and conditions, determining the next execution direction; execution nodes are used to execute specific operations and tasks. Based on the logical relationships and flow sequence in the decision logic graph, the arrangement order and interrelationships of judgment nodes and execution nodes are determined. Functions for key judgment nodes in the decision reasoning chain are recorded; these functions implement the judgment logic of the judgment nodes. Simultaneously, the required data input interfaces are recorded, clarifying the data types, formats, and content that judgment nodes need to receive, to ensure that judgment nodes can accurately make judgments and decisions.

[0068] Step S1267: Define a structured template for the decision output content for each rule branch path, and record the operation type code list, execution requirement parameter list, resource configuration suggestion vector, and risk prevention and control key point identification core information that the decision result should include.

[0069] A structured template for the decision output is designed for each rule branch path. This template specifies the format and content of the decision result. The template includes: a list of operation type codes (encoding the specific operations recommended by the decision); a list of execution requirement parameters (clarifying the parameters and requirements that must be met when performing these operations); a resource allocation recommendation vector (providing suggestions on resource allocation and configuration); and risk prevention and control point identifiers (indicating potential risks and corresponding prevention and control measures during decision execution). By defining this structured template, the consistency and completeness of the decision output are ensured, facilitating subsequent understanding, execution, and evaluation of the decision results.

[0070] Step S1268: Based on the preset terminology dictionary, perform a standardized replacement operation on all terminology expressions in the initial decision rule set, and replace the identified synonyms with the corresponding standard terms according to the preset terminology mapping table.

[0071] The terminology in the initial decision rule set is standardized using a pre-defined terminology dictionary. This dictionary contains standard terms and related synonyms in the field of forest resource management. All terms in the initial decision rule set are scanned and identified to find synonyms. Based on a pre-defined terminology mapping table, the identified synonyms are replaced with their corresponding standard terms. This standardized replacement process ensures the consistency and accuracy of terminology in the initial decision rule set, avoiding misunderstandings and decision-making errors caused by inconsistent terminology, and improving the readability and executability of the decision rules.

[0072] Step S1269: Construct an automatic verification process for the initial decision rule set, record the logical consistency verification algorithm, coverage integrity evaluation function and execution feasibility check program of the rule set, and output the initial decision rule set containing rule structure element data, correlation data of each element and rule application scope definition data.

[0073] An automated validation process is established for the initial decision rule set to ensure its quality and effectiveness. This process includes logical consistency verification, coverage integrity assessment, and execution feasibility checks. The logical consistency verification algorithm checks for logical contradictions and conflicts between rules in the set, ensuring correct logical relationships. The coverage integrity assessment function evaluates whether the rule set covers all possible decision scenarios and situations, ensuring no important decision points are omitted. The execution feasibility check analyzes the feasibility of decisions and operations in the rule set during actual execution, considering limitations such as resources, technology, and environment. The automated validation process comprehensively checks and evaluates the initial decision rule set, and adjustments and optimizations are made based on the results. The final output is an initial decision rule set containing rule structure element data, relationship data between elements, and rule application scope definition data.

[0074] Step S127: Perform condition matching and fusion between the initial decision rule set and the scenario constraint data list in the scenario requirement analysis result, modify the execution constraint data in the rule set to make the constraint data in the rule set consistent with the current scenario constraint data list, and form a preliminary decision adaptation rule set.

[0075] The initial set of decision rules is matched against the list of scenario constraints in the scenario requirements analysis results. The execution constraints in the initial set of decision rules are compared one by one with those in the scenario constraint list to identify differences and inconsistencies. Based on the content of the scenario constraint list, the execution constraints in the initial set of decision rules are modified and adjusted to ensure consistency with the constraints of the current scenario. Through condition matching and fusion operations, the generated preliminary set of decision-adaptive rules is ensured to adapt to the specific constraints of the current scenario, improving the applicability and relevance of the decision rules.

[0076] Step S128: Perform logical consistency verification on the preliminary decision adaptation rule set, compare the logical relationships between the clauses within the rules, mark the clauses with logical conflicts as pending processing, and generate replacement clauses to update the original conflicting clauses based on the preset conflict resolution rules or by calling the knowledge base; for clauses with missing information, generate supplementary content based on the related information in the knowledge base to obtain the improved preliminary decision adaptation rule set.

[0077] The initial decision-making adaptation rule set undergoes logical consistency verification. The logical relationships between the clauses in the rule set are analyzed to check for contradictory or conflicting clauses. If a logically conflicting clause is found, it is marked as pending. Conflicting clauses are processed according to pre-defined conflict resolution rules, such as priority rules and majority rules; alternatively, relevant knowledge and rules from the knowledge base are invoked to generate replacement clauses that resolve the conflict, updating the original conflicting clauses. Simultaneously, the rule set is checked for missing information. For these clauses, supplementary content is generated based on related information in the knowledge base, improving the clause information. Through logical consistency verification and information supplementation, a refined initial decision-making adaptation rule set is obtained, enhancing the logicality and completeness of the rule set.

[0078] Step S129: Verify the improved preliminary decision adaptation rule set with the basic normative document library of natural resource management, compare the differences between the rule clauses and the core requirements of the basic norms, update the inconsistent clauses according to the core requirements of the basic norms, and output the large model decision adaptation rules containing a data table of the mapping relationship between needs and decisions and a list of execution constraints.

[0079] The refined preliminary decision-making adaptation rule set was verified against the basic normative document library for natural resource management. Each clause in the rule set was compared with the core requirements of the basic normative documents in the library to identify differences and inconsistencies. Based on the core requirements of the basic normative documents, the inconsistent clauses in the rule set were updated and modified to ensure compliance. After verification and updating, the large-scale model decision-making adaptation rules were output, which included a data table mapping the relationship between needs and decisions and a list of execution constraints.

[0080] Step S130: Drive the pre-trained natural resource management big model to output a management action combination scheme through the big model decision adaptation rules. Input the big model decision adaptation rules and the natural resource current status data stream obtained from the monitoring terminal into the decision generation module of the big model for processing to generate a management action combination scheme. The management action combination scheme includes a specific management operation instruction sequence and an execution timing schedule table.

[0081] The pre-trained large-scale natural resource management model receives decision adaptation rules and current state data streams of natural resources acquired from monitoring terminals as input. Monitoring terminals collect real-time data on the current state of natural resources, such as tree growth, soil fertility, and pest and disease occurrence in forest areas, forming a data stream that is input into the large-scale model. The decision generation module of the large-scale model analyzes and processes the current state data stream of natural resources according to the decision adaptation rules. Utilizing the model's internal algorithms and logic, combined with the mapping relationships and constraints in the decision adaptation rules, the decision generation module generates specific management action combinations. These management action combinations include a sequence of specific management operation instructions, i.e., the order and content of the specific management operations to be performed; and an execution timing schedule, specifying the execution time, sequence, and time intervals for each management operation. Through the processing of the large-scale model, it is ensured that the generated management action combinations can effectively achieve management objectives based on the current state of natural resources and the decision rules.

[0082] Step S131: Load the pre-trained natural resource management big model, which includes a data preprocessing module, a rule parsing module, a decision generation module, and a result optimization module.

[0083] The pre-trained large-scale natural resource management model is launched and loaded. This model consists of multiple functional modules, including a data preprocessing module, a rule parsing module, a decision generation module, and a result optimization module. The data preprocessing module is responsible for cleaning, transforming, and standardizing the input data to ensure data quality and consistency. The rule parsing module parses the decision adaptation rules of the large-scale model, converting them into an internal representation that the model can understand and execute, and extracting the mapping relationships and constraints. The decision generation module is the core module of the large-scale model. Based on the preprocessed data and parsed decision rules, it uses deep learning algorithms and decision logic to generate management action combination schemes. The result optimization module optimizes and adjusts the generated management action combination schemes, considering the influence of various factors to improve the feasibility and effectiveness of the schemes. The modules interact and collaborate through a data flow interface to jointly complete the generation process of management action combination schemes.

[0084] Step S132: Input the current state data stream of natural resources into the data preprocessing module of the large model for missing value imputation, outlier processing and standardization transformation to form a standardized state data matrix. The current state data stream of natural resources includes a real-time monitoring data set consisting of resource quantity time series data, resource distribution spatial data, resource quality index data and resource surrounding environmental parameter data.

[0085] The current status data stream of natural resources obtained from the monitoring terminal will be input into the data preprocessing module of the large model. The current status data stream of natural resources contains various types of data, such as resource quantity time series data, that is, data on changes in resource quantity recorded at different time points; resource distribution spatial data, reflecting the spatial distribution of resources; resource quality index data, measuring the quality status of resources; and environmental parameter data of the surrounding environment of resources, such as environmental data that affect resource growth, such as temperature, humidity, and light.

[0086] The data preprocessing module processes the aforementioned data. First, missing value imputation is performed. For missing values ​​in the data, appropriate methods are used to impute them based on the data's characteristics and distribution, such as mean imputation, median imputation, and interpolation, ensuring data integrity. Next, outlier handling is performed. Statistical analysis methods are used to identify outliers in the data, such as standard deviation analysis and box plot methods, to correct or remove outliers and prevent them from interfering with subsequent processing. Finally, standardization transformation is performed, converting data of different magnitudes and units into a unified standard format, such as converting the data to a standard normal distribution with a mean of 0 and a standard deviation of 1, forming a standardized state data matrix. Standardization helps eliminate the influence of data dimensions, improving the accuracy and efficiency of the model in processing data.

[0087] Step S133: Input the large model decision adaptation rules into the rule parsing module of the large model, perform syntax parsing and semantic parsing on the data table of the mapping relationship between demand and decision and the list of execution constraints in the large model decision adaptation rules, extract the feature vectors of key decision elements and execution condition predicates in the rules, and form the rule parsing result data structure.

[0088] The decision adaptation rules for the large model are input into the rule parsing module of the large model. The rule parsing module performs syntactic and semantic parsing on the decision adaptation rules. During syntactic parsing, the correctness of the syntactic structure of the decision adaptation rules is checked to ensure that the format of the rules meets the model's requirements. Semantic parsing delves into the meaning of the rules, analyzing the mapping relationship data table between requirements and decisions and the list of execution constraints. The correspondence between requirements and decisions is extracted from the mapping relationship data table, and the specific content of each constraint is extracted from the list of execution constraints.

[0089] By parsing, feature vectors of key decision elements in the rules are extracted. These feature vectors can reflect the key factors and characteristics of the decision. At the same time, execution condition predicates are extracted to clarify the conditions that need to be met for the execution of the decision.

[0090] Step S134: Input the standardized state data matrix and the rule parsing result data structure into the decision generation module of the large model. The decision generation module performs feature matching in the standardized state data matrix based on the key decision element feature vector in the rule parsing result data structure, and searches in the preset management operation knowledge base. The management operation types with a matching degree higher than the set threshold are encoded and output as the filtering results.

[0091] The standardized state data matrix generated by the data preprocessing module and the rule parsing result data structure output by the rule parsing module are input into the decision generation module of the large model. The decision generation module first performs feature matching in the standardized state data matrix based on the key decision element feature vectors in the rule parsing result data structure. By calculating the similarity between feature vectors, it identifies state data features related to the key decision element feature vectors.

[0092] Simultaneously, a search is performed in a pre-defined management operation knowledge base. This knowledge base stores various possible management operation types and their related characteristics. Based on the matched status data features and key decision element feature vectors, the system searches the knowledge base for matching management operation types. A matching threshold is set, and management operation types with a matching degree higher than this threshold are encoded and output as the filtering results. The filtering results contain management operation types that initially meet the criteria.

[0093] Step S135: Based on the execution condition predicate in the rule parsing result data structure, solve the sorting constraints for the selected management operation type codes, calculate the mutual influence and synergy measures between different management operation type code combinations, arrange the execution sequence of various operation codes, and form a preliminary operation sequence scheduling framework.

[0094] Based on the execution condition predicates in the rule parsing result data structure, the sorting constraints of the management operation type codes selected in step S134 are solved. The execution condition predicates specify the conditions and order of execution of management operations. During the sorting constraint solution process, these constraints are considered to determine the possible execution order of the management operation type codes.

[0095] The interaction metrics between different management operation type code combinations are calculated to analyze the potential promoting or inhibiting effects between different operations; simultaneously, synergy metrics are calculated to evaluate the overall effect produced when different operation combinations are executed together. Based on the interaction metrics, synergy metrics, and execution condition predicates, the execution sequence of various operation codes is arranged. The arranged execution sequence is organized into a preliminary operation timing scheduling framework, which specifies the approximate execution order and time arrangement of each management operation.

[0096] Step S1351: Extract the list of execution condition predicates from the rule parsing result data structure, parse out the execution prerequisite predicates, execution time interval predicates, and execution order predicates for various management operations, and organize all the parsed execution condition predicates into a list of constraint condition predicates.

[0097] The execution condition predicate list is extracted from the rule parsing result data structure. This list is then parsed in detail to identify the execution prerequisite predicates for various management operations (i.e., the conditions that must be met for the operation to execute); the execution time interval predicate, which specifies the time range for operation execution; and the execution order predicate, which clarifies the execution order relationship between different operations. All the parsed execution condition predicates are then organized to form a constraint condition predicate list.

[0098] Step S1352: Extract the functional descriptions and attribute parameters corresponding to the selected management operation type codes, record the implementation cycle parameters, the list of resource types required for implementation, the scope of implementation impact, and the description of subsequent impacts after implementation for each type of management operation, and form a list of operation feature parameters.

[0099] For each selected management operation type code, its corresponding functional description and attribute parameters are extracted from the management operation knowledge base. The functional description explains the specific function and purpose of the management operation; the attribute parameters include the implementation cycle parameter, i.e., the time required to complete the operation; a list of required resource types, listing the various resources needed to execute the operation; the spatial scope of the implementation impact, describing the area of ​​spatial impact; and a description of the subsequent impacts after implementation, explaining the potential impacts on other aspects after the operation is executed. This information is recorded to form a list of operation characteristic parameters.

[0100] Step S1353: Based on the list of constraint predicates and the list of operation feature parameters, use a directed graph construction algorithm to identify the dependencies between various management operations, mark management operations with subsequent dependencies, management operations that can be executed in parallel, and management operations with mutual exclusion, and generate a directed graph of operation dependencies.

[0101] By combining a list of constraint predicates and a list of operation feature parameters, a directed graph construction algorithm is used to build a dependency model among management operations. Nodes in the directed graph represent the type encoding of the management operation, and directed edges represent the dependencies between operations. Based on information such as the execution cycle and subsequent impact in the execution order predicates and operation feature parameters, management operations with post-dependencies are identified (i.e., the execution of one operation must be after the completion of another); management operations that can be executed in parallel (i.e., multiple operations can be performed simultaneously without affecting each other); and management operations with mutual exclusion (i.e., two operations cannot be executed simultaneously, and the execution of one will affect the effect or feasibility of the other). These relationships are then marked in the directed graph to generate a directed graph of operation dependencies.

[0102] Step S1354: Calculate the management effectiveness prediction data corresponding to different management operation combination sequences, call the predefined synergy effect calculation function, calculate the synergy effect value of different management operation combinations, place the combinations with positive synergy effect values ​​at the front in the sorting sequence, and avoid management operation sequences that may generate negative mutual influence based on the predefined conflict detection model.

[0103] For different combinations of management operations, a predefined management effectiveness prediction model is used to calculate corresponding management effectiveness prediction data, which reflects the expected management results under that combination sequence. Simultaneously, a synergy effect calculation function is invoked to calculate the synergy effect values ​​for different combinations of management operations. A positive synergy effect value indicates that the operation combination can produce a mutually reinforcing effect; the larger the value, the more significant the reinforcing effect. When ranking operations, combinations with positive synergy effect values ​​are placed at the beginning of the ranking sequence to prioritize operation sequences that can produce synergistic effects. Furthermore, based on a predefined conflict detection model, potential negative interactions in the management operation sequences, such as resource competition and effect cancellation, are analyzed. Management operation sequences that may produce negative interactions are avoided to ensure the rationality and effectiveness of the ranking results.

[0104] Step S1355: Combining the overall target direction data list of natural resource management, calculate the execution priority weight of management operations according to the contribution calculation model of management operations to target items. Operations with a contribution weight higher than the preset threshold are marked as priority execution level and assigned higher priority during sorting.

[0105] Based on the overall target and direction data list of natural resource management, the importance and priority of each management objective are clarified. According to the contribution of management operations to each objective, a contribution calculation model is used to calculate the execution priority weight of management operations. The contribution calculation model considers factors such as the direct impact, indirect impact, and degree of impact of the operation on the objective, assigning a priority weight value to each management operation. Management operations with a contribution weight higher than a preset threshold are marked as priority execution levels. When ranking operations, these priority execution levels are assigned higher priority to ensure they are executed first, thus better achieving the overall management objectives.

[0106] Step S1356: Based on the directed graph of operation dependencies, synergy measurement data, and operation execution priority weights, take the directed graph of operation dependencies, synergy measurement data, and operation execution priority weights as inputs, call the time-series planning algorithm based on constraint satisfaction to perform calculations, output the execution sequence of various management operations, and record the planned execution stage code and planned time window for each management operation.

[0107] The algorithm takes a directed graph of operation dependencies, synergy metrics, and operation execution priority weights as input and processes them using a constraint-based temporal programming algorithm. The temporal programming algorithm comprehensively considers the dependencies, synergies, and execution priorities between operations to determine the optimal execution sequence for various management operations while satisfying various constraints. The algorithm outputs the execution sequence of management operations and records a planned execution stage code for each operation to identify its position in the overall execution process; as well as a planned time window, specifying the start and end time ranges of the operation. Through the calculation of the temporal programming algorithm, a reasonable execution order and time schedule are generated.

[0108] Step S1357: Compare the proposed execution sequence with the list of constraint predicates, check for violations of execution constraints through the constraint satisfaction problem solver, and identify problems such as resource supply conflicts or overlapping time windows.

[0109] The proposed execution sequence generated by the time-series programming algorithm is compared with the list of constraint predicates. A constraint satisfaction problem solver is used to check the execution sequence to verify whether it meets all execution constraints in the constraint predicate list. Special attention is paid to investigating resource supply conflicts (where multiple operations simultaneously demand the same resource beyond its supply capacity) and overlapping time windows (where the planned time windows of different operations overlap, potentially leading to execution conflicts). Through this check, violations of constraints and other problematic points in the execution sequence are identified.

[0110] Step S1358: For the identified problems, adjust the execution sequence of relevant management operations, adjust the start and end nodes of the planned time window for each management operation, recalculate and allocate resource supply quotas, and re-execute the constraint satisfaction problem solving for verification.

[0111] For identified issues such as resource supply conflicts and overlapping time windows, the execution sequence of related management operations is adjusted. The execution order of management operations is rearranged based on the severity and scope of the problem. Simultaneously, the start and end points of the planned time windows for each management operation are adjusted to avoid overlapping time windows. For resource supply conflicts, resource supply quotas are recalculated and reallocated to ensure that each operation receives sufficient resource support during execution. After adjustment, the constraint satisfaction problem solver is re-executed for verification to check whether the problem points have been resolved and whether the execution sequence satisfies all constraints. If problems persist, the adjustment and verification process is repeated until all problem points are resolved.

[0112] Step S1359: Connect to the professional knowledge graph database in the field of natural resource management, evaluate the adjusted execution sequence using the entity relationships in the knowledge graph, and adjust the management operation timing arrangement with reference to the standard process entity and practical case entity data in the professional knowledge graph.

[0113] This system connects to a professional knowledge graph database in the field of natural resource management, which contains rich domain knowledge and entity relationships. Entity relationships within the knowledge graph, such as those between management operations and entities related to objectives, resources, and the environment, are used to evaluate the adjusted execution sequence. Referring to standard process entities in the knowledge graph—recognized standard operating procedures within the domain—and case study entity data—the operational sequence arrangements from successful past practices—the current management operation sequence is further adjusted and optimized. By combining the professional knowledge graph with this system, the management operation sequence arrangement becomes more aligned with domain expertise and practical experience, improving its rationality and reliability.

[0114] Step S13510: Based on the adjusted management operation execution sequence, planned execution phase code, and planned time window, a preliminary operation timing scheduling framework is formed.

[0115] The adjusted and optimized sequence of management operations execution, the planned execution phase code for each management operation, and the planned time window are integrated. Management operations are grouped according to the planned execution phase code, clarifying the operational content included in each phase. Within each phase, the execution order and timing of operations are arranged according to the planned time window. This information is organized into a structured preliminary operation timing framework, which demonstrates the execution order, phase division, and time arrangement of management operations.

[0116] Step S136: Based on the preliminary operation timing scheduling framework, retrieve the detailed operation template corresponding to each management operation type code from the management operation knowledge base, fill the corresponding parameter values ​​in the standardized state data matrix into the corresponding variable positions of the detailed operation template, and generate detailed operation plan data containing operation implementation object identifier, implementation method code, implementation scope coordinates and implementation resource list.

[0117] Based on the management operation type codes determined in the preliminary operation timing scheduling framework, detailed operation templates corresponding to each operation type code are retrieved from the management operation knowledge base. Each detailed operation template contains the specific steps, methods, and required parameters for operation implementation. It includes multiple variable positions, which need to be filled with specific parameter values ​​according to the actual situation. Parameter values ​​related to the operation are extracted from the standardized state data matrix and filled into the corresponding variable positions in the detailed operation template. These include: operation implementation object identifier (identifying the specific object of the operation); implementation method code (specifying the implementation method); implementation scope coordinates (determining the spatial scope of the operation); and a list of required resources (listing the types and quantities of resources needed to execute the operation). By filling in the parameter values, detailed operation plan data is generated, making management operations more specific and executable.

[0118] Step S137: Input the detailed operation plan data into the result optimization module of the large model. The result optimization module combines the operation effect index data in the historical decision case database, obtains the effect index of similar historical operations under similar parameters from the operation effect index database, compares it with the expected effect of the current operation parameters, updates and calculates the current operation parameters based on the comparison results, and generates detailed operation plan data after parameter adjustment.

[0119] The generated detailed operational plan data is input into the result optimization module of the large model. The result optimization module accesses the operational performance indicator database in the historical decision case database, which stores performance indicator data for various management operations under different parameter conditions. For each management operation in the current detailed operational plan data, performance indicator data for similar historical operations under similar parameter conditions is retrieved from the operational performance indicator database. This historical performance indicator data is compared and analyzed with the expected results under the current operational parameters to identify the differences. Based on the comparison results, the current operational parameters are updated and calculated using optimization algorithms, adjusting parameter values ​​to improve the expected operational results. The adjusted detailed operational plan data is then generated, further optimizing the management operational plan and improving the reliability of its execution.

[0120] For example, step S1371: retrieve the operation effect index data records from the historical decision case database. The operation effect index data records include the implementation effect evaluation indicators of various management operations, the effect influencing factor analysis report, and the operation optimization suggestion list. The operation effect index data records are classified, stored, and indexed according to the operation type code to form an operation effect index database.

[0121] Operational effectiveness indicator data records are retrieved from the historical decision-making case database. These records include evaluation indicators for the implementation effects of various management operations, such as the percentage increase in forest coverage, tree survival rate, and pest and disease control rate. An analysis report of influencing factors analyzes various factors affecting operational effectiveness. A list of operational optimization suggestions provides recommendations for different situations. The operational effectiveness indicator data records are categorized and stored according to management operation type codes, with an index created for each type of operation to facilitate quick querying and retrieval. Through categorized storage and indexing, an operational effectiveness indicator database is formed.

[0122] Step S1372: Match and retrieve the various management operation codes in the detailed operation plan data with the same type of management operation codes in the operation effect index database, extract the historical effect evaluation index set and influencing factor analysis report of the matching management operations, and record the effect performance index data of the corresponding management operations under different implementation condition parameters.

[0123] The system matches and retrieves the various management operation codes from the detailed operational plan data with the management operation type codes from the operational performance indicator database. For each successfully matched management operation, its corresponding historical performance evaluation indicator set is extracted. This set contains the performance indicator values ​​of the operation in different historical cases. Simultaneously, an influencing factor analysis report is extracted to identify the main factors affecting the operation's effectiveness. The system records the performance indicator data of the corresponding management operation under different implementation conditions, such as changes in performance indicators under different resource inputs, environmental conditions, and execution time parameters.

[0124] Step S1373: Compare the implementation condition parameters of each management operation in the detailed operation plan data with the historical implementation condition parameters of similar management operations in historical decision-making cases, extract a list of key difference parameters, and evaluate the potential impact weight of key difference parameters on the operation effect.

[0125] The implementation condition parameters for each management operation in the detailed operational plan data are compared with the historical implementation condition parameters for similar management operations in historical decision-making cases. Differences are identified, and a list of key difference parameters is extracted. These parameters are those that may significantly impact the operational outcome. By analyzing historical data and influencing factor analysis reports, the potential impact weight of each key difference parameter on the operational outcome is assessed. The higher the impact weight, the more significant the impact of the parameter's difference on the operational outcome. By evaluating the impact weight of key difference parameters, current operational parameters can be adjusted more effectively to reduce the negative impact of differences and improve operational results.

[0126] Step S1374: Refer to the list of optimization suggestions for similar management operations in historical decision-making cases, and combine the implementation condition parameters and scenario requirement analysis results of the current detailed operation plan data to generate a parameter adjustment instruction set for each management operation, and record the key parameters and expected optimization target values ​​of the adjustment instruction set.

[0127] Referencing a list of optimization suggestions for similar management operations from historical decision-making cases, these suggestions, based on historical experience and data analysis, possess significant reference value. Combining the implementation condition parameters and scenario requirement analysis results of the current detailed operational plan data, we analyze the existing problems with the current operational parameters and the directions for optimization. Based on historical optimization suggestions and the current situation, we generate parameter adjustment instruction sets for each management operation. These instruction sets clearly specify the names of the parameters to be adjusted, the direction of adjustment, and the magnitude of adjustment. Simultaneously, we record the key parameters in the adjustment instruction sets—those critical parameters significantly impacting the operational effectiveness—as well as the expected optimization target values—that are, the parameter values ​​and performance indicators desired after the adjustments.

[0128] Step S1375: Adjust the implementation code for each management operation, and modify the implementation step sequence, implementation tool selection, or implementation parameter configuration of the management operation based on the effect feedback in the operation effect index data record.

[0129] Based on the feedback information recorded in the operational performance metrics data, analyze whether the implementation code of the current management operation is appropriate. If the feedback indicates that the current implementation method has problems or room for optimization, adjust the implementation code. Modify the sequence of implementation steps of the management operation to optimize the operation process; adjust the selection of implementation tools to choose tools more suitable for the current conditions; or modify the implementation parameter configuration, such as adjusting parameters such as operation intensity and frequency. By adjusting the implementation code, improve the adaptability and effectiveness of the management operation to achieve better operational results.

[0130] Step S1376: Adjust the coordinates of the implementation scope of management operations. Combine the correlation analysis model of operation scope and effect in historical decision-making cases to calculate and adjust the implementation space boundary of some management operations.

[0131] This paper utilizes a correlation analysis model based on historical decision-making cases to illustrate the relationship between the scope and effects of management operations. Based on current scenario requirements and detailed operational plan data, the paper analyzes the rationality of the current management operation's implementation scope coordinates. The model calculates the expected effects under different implementation spatial boundaries, and adjusts the implementation spatial boundaries of some management operations based on the calculation results. The implementation scope is expanded or reduced to maximize operational effectiveness while avoiding unnecessary resource waste and impact on non-target areas.

[0132] Step S1377: Adjust the resource allocation ratio vector and resource supply method of various management operations in the detailed operation plan data, and refer to the resource consumption index data and effect index data of similar management operations in historical decision-making cases.

[0133] By referencing resource consumption and effectiveness data from similar management operations in historical decision-making cases, the relationship between resource consumption and effectiveness is analyzed. Based on the resource requirements and supply capabilities of various management operations in the current detailed operational plan data, the resource allocation ratio vector is adjusted, i.e., the allocation ratio of various resources among different operations. Simultaneously, resource supply methods are adjusted, such as selecting more efficient resource supply channels and optimizing resource scheduling methods. Through reasonable adjustments to resource allocation and supply methods, the effective utilization of resources is ensured, improving the overall effectiveness of management operations.

[0134] Step S1378: Compare the synergy measurement of various management operations in the adjusted detailed operation plan data, and correct the points of cooperation conflict or duplicate content identification between management operations.

[0135] The synergy metrics of various management operations in the adjusted detailed operational plan data were recalculated and compared. Analysis was conducted to identify any points of conflict between management operations, i.e., the execution of some operations might negatively impact the effectiveness of others; or duplicate content markers, i.e., overlapping work content exists between different operations. To address these issues, management operations were revised, adjusting parameters, timing, or content to eliminate points of conflict and duplicate content, thereby improving the synergy between management operations and ensuring the overall efficiency and consistency of the operational plan.

[0136] Step S1379: Construct a simulation environment based on historical decision-making case data, perform discrete event simulation on the adjusted detailed operation plan, execute the plan simulation process, record the management operation simulation implementation effect index data and resource consumption simulation data, and identify a list of potential problem points.

[0137] A simulation environment is constructed based on historical decision-making case data, capable of simulating various situations and changes in natural resource management scenarios. The adjusted detailed operational plan is input into the simulation environment for discrete event simulation. During the simulation, the implementation process of management operations is simulated according to the execution sequence and content. Data on the simulated implementation effect of management operations is recorded, such as simulated changes in forest cover and the degree of improvement in tree growth; as well as simulated resource consumption data, such as water and fertilizer resources consumed during the simulation. Through simulation, a list of potential problems in the detailed operational plan is identified, such as insufficient resources, poor results, and operational conflicts.

[0138] Step S13710: Based on the list of potential problem points recorded in the simulation, adjust the detailed parameters of the management operation to form optimized detailed operation plan data.

[0139] Based on the list of potential problem points identified during the simulation, the detailed parameters of the management operations were adjusted. For each problem point, its cause was analyzed, and corresponding operational parameters were adjusted, such as modifying the implementation time, adjusting resource allocation, or changing the operation method. By adjusting the detailed parameters, potential problem points were resolved, improving the feasibility and effectiveness of the detailed operational plan data. After multiple simulations and adjustments, optimized detailed operational plan data was generated.

[0140] Step S138: Analyze the resource consumption estimates for various operations in the detailed operation plan data, and combine them with the resource supply capacity data of the current natural resource management scenario to recalculate and allocate the resource allocation list in the operation plan.

[0141] Analyze the resource consumption estimates for various operations within the detailed operational plan data to understand the resource requirements of each operation. Simultaneously, acquire resource supply capacity data for the current natural resource management scenario, including the available quantity and supply rate of various resources. Compare the resource consumption estimates with the resource supply capacity data to assess whether the resource supply can meet the operational needs. If there are insufficient resource supplies or unreasonable allocations, recalculate and reallocate the resource allocation list in the operational plan. Adjust the resource allocation ratios between different operations and optimize the resource supply methods to ensure that all operations receive sufficient resource support during execution, avoiding disruptions to normal operation and effectiveness due to resource issues.

[0142] Step S139: Perform a global constraint check on the adjusted detailed operation plan data, detect time windows that violate preset constraints in the operation execution timing schedule, recalculate and allocate time windows; identify and mark logical conflicts between operation content data, and select or sort conflicting operations according to preset priority rules; check the demand and supply of each resource in the resource configuration list, recalculate and allocate data items with supply and demand mismatch, and form a management action combination plan that includes a specific management operation instruction sequence and an execution timing schedule table.

[0143] A global constraint check is performed on the adjusted detailed operation plan data to ensure that all aspects of the plan meet the preset constraints. Regarding operation execution timing, it checks for time windows that violate preset constraints, such as operation execution times exceeding the allowed time range or overlapping time windows. Time windows are then recalculated and reassigned to ensure the rationality of the timing schedule. Logical conflicts between operation content data are identified and marked, such as contradictory operation objectives or incompatible operation methods. Conflicting operations are selected or prioritized according to preset priority rules, such as objective priority and effect priority, to eliminate logical conflicts. The demand and supply of each resource in the resource configuration list are checked, and data items with supply-demand mismatches are identified. These are then recalculated and reassigned to ensure resource supply-demand balance. After global constraint checks and adjustments, the final management action combination plan is formed, which includes a specific management operation instruction sequence and an execution timing schedule table.

[0144] Step S140: The management action combination scheme is issued to the execution unit in the natural resource management scenario. The resource status change data stream, operation execution effect log and scenario environment change information during the scheme execution process are collected by the sensor network and execution log collection system deployed in the scenario. After cleaning, fusion and formatting, a multi-dimensional feedback dataset is generated. The multi-dimensional feedback dataset includes a set of execution effect indicators and a set of scenario change features.

[0145] The generated management action combination plan is distributed to the execution units in the natural resource management scenario. The execution units then execute the corresponding management operations according to the specific management operation instruction sequence and execution timing schedule in the plan. During the execution of the plan, a sensor network deployed in the scenario collects real-time data streams on changes in resource status, such as changes in the growth status of trees in forest areas, changes in soil fertility, and changes in the occurrence of pests and diseases. Simultaneously, the execution log collection system records operation execution effect logs, including the execution status of operations and the achievement of various indicators. In addition, information on changes in the scenario environment, such as changes in weather, policy adjustments, and other external environmental factors, is collected.

[0146] The collected resource status change data stream, operation execution effect logs, and scene environment change information are processed. First, a data cleaning operation is performed to remove noise, outliers, and missing values, ensuring data accuracy and completeness. Then, a fusion operation is performed to integrate data from different sources and of different types, establishing relationships between the data. Finally, formatting is performed to convert the data into a unified format for easier subsequent analysis and use. The processed dataset generates a multi-dimensional feedback dataset, which includes a set of execution effect indicators reflecting the various metrics of operation execution effectiveness, and a set of scene change features reflecting changes in the scene environment.

[0147] Step S141: Deploy a distributed data acquisition network. The distributed data acquisition network covers the key areas and operation execution areas of the natural resource management scenario. The deployment coordinates and acquisition range parameters of the acquisition nodes are determined based on the operation implementation space range and resource distribution heat map data in the management action combination scheme.

[0148] In natural resource management scenarios, a distributed data acquisition network is deployed to cover key areas such as ecologically sensitive areas and resource-concentrated areas, as well as operational execution areas, i.e., the operational implementation areas specified in the management action combination scheme. The deployment coordinates and acquisition range parameters of the acquisition nodes are determined based on the operational implementation spatial range and resource distribution heatmap data in the management action combination scheme. The resource distribution heatmap data reflects the spatial distribution density and importance of resources. Based on this data, the deployment density of acquisition nodes is increased in densely distributed and important resource areas to expand the acquisition range, ensuring comprehensive and accurate collection of data on resource status changes and operational execution effects. The acquisition nodes employ various types of sensors, such as temperature sensors, humidity sensors, and image sensors, to meet the acquisition needs of different types of data.

[0149] Step S142: Configure the data acquisition frequency parameters and the list of acquisition content identifiers for the data acquisition nodes. The acquisition frequency parameters are determined based on the implementation cycle parameters of the management operation and the resource status change rate model. The list of acquisition content identifiers includes resource status change information composed of time-series data on the quantity change of natural resources, quality change index data, and spatial data on distribution change.

[0150] Each data acquisition node in the distributed data acquisition network is configured with parameters, including acquisition frequency parameters and a list of acquisition content identifiers. The acquisition frequency parameter is determined based on the implementation cycle of management operations and the resource status change rate model. Higher acquisition frequencies are set for areas with shorter implementation cycles or faster resource status changes to ensure timely capture of data changes; conversely, lower acquisition frequencies are set to reduce unnecessary data acquisition and transmission. The list of acquisition content identifiers clarifies the data content that each acquisition node needs to collect, including time-series data on changes in the quantity of natural resources, such as increases or decreases in the number of trees; quality change index data, such as changes in timber quality and soil fertility; and spatial data on distribution changes, such as resource status change information including spatial movement and changes in distribution range. By configuring the acquisition parameters and content, it is ensured that the acquisition nodes can collect effective data as needed.

[0151] Step S143: Collect execution effect log data of management operations, including resource response index data after operation implementation, target achievement progress percentage data, resource consumption list during operation implementation, and data stream reflecting the actual effect of operation, which consists of operation execution smoothness index.

[0152] The execution log collection system is specifically designed to collect execution effect log data for management operations. This execution effect log data includes resource response indicators after the operation is implemented, i.e., the resource's reaction and changes in response to the operation; target achievement percentage data, measuring the completion of the operation's objectives; a resource consumption list during the operation implementation process, recording the quantity of various resources consumed during the operation; and operation execution smoothness indicators, such as whether the operation proceeded smoothly as planned, whether any faults or delays occurred, etc. All of the above data together constitute a data stream reflecting the actual effect of the operation, which is collected in real time by the execution log collection system and transmitted to the data processing center.

[0153] Step S144: Collect scene environment change information stream, including remote sensing monitoring data of the natural environment, text update stream of policy requirements, and external environment information affecting natural resource management composed of parameter change information of technical application conditions.

[0154] Information on changes in the environment is collected through various channels. Remote sensing data of the natural environment is acquired through technologies such as satellite and aerial remote sensing, including land use change, vegetation cover change, and meteorological data. Textual updates related to policy requirements are obtained by regularly retrieving policy documents, regulations, and ordinances issued by government departments to promptly understand policy changes. Information on changes in parameters of technological application conditions is obtained by monitoring the performance parameters of technical equipment and software version updates. This information constitutes the external environmental information affecting natural resource management. Collecting this information helps to adjust management strategies and decision-making plans in a timely manner to adapt to changes in the external environment.

[0155] Step S145: Transmit the collected resource status change data stream, operation execution effect log data stream, and scene environment change information stream in real time, and transmit all kinds of raw collected data to the data processing center server through the preset data transmission protocol.

[0156] The collected raw data, such as resource status change data streams, operation execution effect log data streams, and scene environment change information streams, are transmitted in real time through a preset data transmission protocol. Common protocols such as TCP / IP, MQTT, and HTTP can be used to ensure the reliability and real-time performance of data transmission. During transmission, the data is encrypted to prevent data leakage and tampering. The data is then transmitted to a data processing center server for centralized processing and storage.

[0157] Step S146: Classify and store the raw collected data transmitted to the data processing center server. According to the data pattern, the raw data is divided into resource status data table, execution effect data table and environmental change data table. Each type of data table is then mapped and stored according to the predefined data pattern.

[0158] The data processing center server classifies and stores the received raw data. Based on data type and characteristics, the raw data is divided into resource status data tables, execution effect data tables, and environmental change data tables. The resource status data table stores data related to resource status changes; the execution effect data table stores log data of management operations; and the environmental change data table stores information on changes in the scenario environment. Each type of data table is then mapped to its fields according to a predefined data schema, ensuring accurate data storage. Relational or non-relational databases are used for data storage; the appropriate database type is selected based on the characteristics of the data to improve storage efficiency and query performance.

[0159] Step S147: Clean the categorized and stored data to obtain valid data records. Call the preset standardization transformation rules to transform the field names, data formats, units of measurement, and codes of the valid data records to make them conform to the predefined standard data pattern and obtain standardized data records. Then, associate and integrate the standardized data records based on timestamps and spatial coordinates to form a multi-dimensional feedback dataset containing a set of execution effect indicators and a set of scene change features.

[0160] The categorized and stored data undergoes a cleaning process. Data cleaning algorithms remove noise, outliers, and duplicate data, and handle missing values ​​to obtain valid data records. Pre-defined standardization transformation rules are applied to unify the field names of the valid data records, convert data formats (e.g., date format, numeric format), standardize units of measurement (e.g., convert length data from different units to a unified metric unit), and standardize the encoding to conform to a predefined standard data pattern. After obtaining standardized data records, they are associated and integrated based on timestamps and spatial coordinates. Data with the same timestamps and spatial coordinates are merged to establish spatiotemporal relationships between the data. Finally, a multi-dimensional feedback dataset is formed, containing a set of execution performance indicators and a set of scenario change characteristics. This multi-dimensional feedback dataset reflects the execution effectiveness of management operations and changes in the scenario environment.

[0161] Step S150: Iteratively optimize the large model decision adaptation rules and the management action combination scheme using the multi-dimensional feedback dataset. Input the multi-dimensional feedback dataset, the historical version of the large model decision adaptation rules, and the historical version of the management action combination scheme into the iterative optimization algorithm module for processing to generate the optimized decision scheme data package.

[0162] The multi-dimensional feedback dataset, historical versions of large-scale model decision-making adaptation rules, and historical versions of management action combination schemes are input into the iterative optimization algorithm module. This module analyzes and processes the input data, evaluating the effectiveness and existing problems of the current large-scale model decision-making adaptation rules and management action combination schemes during actual execution. Based on the set of execution effect indicators and scenario change characteristics in the multi-dimensional feedback dataset, the deviation between the decision rules and action schemes and the actual situation is analyzed. Iterative optimization algorithms, such as gradient descent and genetic algorithms, are used to adjust the mapping relationships and constraints in the large-scale model decision-making adaptation rules and optimize the operation instruction sequence and execution timing schedule in the management action combination schemes. Through multiple iterations of optimization, the decision rules and action schemes are continuously improved, generating an optimized decision scheme data package. This decision scheme data package contains the optimized large-scale model decision-making adaptation rules and management action combination schemes, which can better adapt to the actual situation of natural resource management scenarios and improve the accuracy and effectiveness of decision-making.

[0163] Step S151: Receive the multi-dimensional feedback dataset, perform feature separation operation on the set of execution effect indicators and the set of scene change features in the multi-dimensional feedback dataset, split the set of execution effect indicators according to the effect dimension label corresponding to the management target direction data list, and divide the set of scene change features according to the change type label and the scope of influence label to form a subset of categorized feedback data.

[0164] The system receives a multi-dimensional feedback dataset and performs feature separation on its set of execution effectiveness indicators and set of scenario change features. For the execution effectiveness indicator set, it is split according to the effectiveness dimension labels corresponding to the management objective direction data list. Each effectiveness dimension label corresponds to a specific management objective, and the execution effectiveness indicator data related to that objective are grouped into one category, forming multiple subsets of execution effectiveness indicators. For the scenario change feature set, it is divided according to change type labels (e.g., natural change, human-caused change, policy change) and impact scope labels (e.g., local impact, regional impact, global impact), forming multiple subsets of scenario change features. Through feature separation, the multi-dimensional feedback dataset is decomposed into multiple categorized feedback data subsets, each with clear features and categories, facilitating subsequent analysis and processing.

[0165] Step S152: Compare the execution effect index data in the subset of classified feedback data with the expected effect index data of the historical version management action combination scheme, calculate the difference value vector, identify the operation records whose effect index difference value exceeds the difference threshold, and record the operation identifier and effect deviation index data of the corresponding management operation.

[0166] The performance metrics data from the categorized feedback data subset are compared with the expected performance metrics data from the historical version management action combination scheme. For each management operation, the difference between the performance metrics data and the expected performance metrics data is calculated, forming a difference value vector. The difference value vector reflects the degree of deviation between the actual performance and the expected performance. A difference threshold is set; when the difference value of the performance metrics exceeds this threshold, the performance of the management operation is considered to have deviated significantly from the expectation. Operation records with performance metrics difference values ​​exceeding the difference threshold are identified, and the operation identifier of the corresponding management operation is recorded to locate the specific operation; at the same time, the performance deviation metrics data, such as the magnitude and direction of the deviation, are recorded.

[0167] Step S153: Based on the effect deviation index data, trace the rule clauses related to the corresponding operation identifiers in the historical version of the large model decision adaptation rules, compare the demand and decision mapping relationship data in the rule clauses with the current scenario feedback data, calculate the adaptation difference value, and locate the rule clause identifiers that cause effect deviation.

[0168] Based on the performance deviation index data, we traced back to the rule clauses related to management operations exhibiting performance deviations in the historical versions of the large model's decision adaptation rules. These rule clauses specified the decision-making basis and constraints for the management operation. We compared the demand-decision mapping data in the rule clauses with the current scenario feedback data to analyze the differences. We calculated the adaptation difference value, which measures the degree of adaptation between the rule clauses and the current scenario; a larger adaptation difference value indicates a higher degree of mismatch between the rule clauses and the current scenario. Based on the adaptation difference value, we located the rule clauses that caused the performance deviations, i.e., identified those rule clauses that did not match the current scenario feedback data, thus leading to performance deviations in management operations.

[0169] Step S154: Compare the scene change feature data in the classification feedback data subset with the rule application scope definition data of the historical version of the large model decision adaptation rules, identify the new change features and new demand features that appear in the current scene, and record the specific data of the new change features and new demand features that are not covered by the historical version of the large model decision adaptation rules.

[0170] The scenario change characteristics data in the subset of categorized feedback data are compared with the rule application scope definition data of the historical version's large-scale model decision adaptation rules. The rule application scope definition data specifies the scenario conditions and scope to which the historical version's decision rules apply. By comparing, new change characteristics appearing in the current scenario, i.e., scenario changes not included in the historical version's rules, and new requirement characteristics, i.e., new management requirements appearing in the current scenario, are identified. Specific data on these new change characteristics and new requirement characteristics not covered by the historical version's large-scale model decision adaptation rules are recorded, such as the type, degree, and scope of impact of the changes.

[0171] Step S155: Based on the identified rule clause identifiers and the specific data of the new change features and new requirement features that are not covered, create new requirement entries and decision mapping relationships corresponding to the new change features and new requirement features in the historical version of the large model decision adaptation rules, and add them as new rule clauses to the mapping relationship data table; identify rule clauses with deviations, and update the decision mapping relationships or condition parameters that do not match the actual scenario feedback data to the corrected content.

[0172] Based on the rule clause identifiers causing effect deviations identified in step S153 and the specific data of uncovered new change features and new requirement features recorded in step S154, the decision adaptation rules of the historical version of the large model are updated and improved. For uncovered new change features and new requirement features, new requirement entries are created in the mapping relationship data table, and corresponding decision mapping relationships are established. These are then added as new rule clauses to the decision adaptation rules to adapt to new scenario changes and requirements. For rule clauses corresponding to rule clause identifiers with deviations, the reasons for their discrepancies with actual scenario feedback data are analyzed, and the decision mapping relationships or condition parameters are updated with corrected content to improve the accuracy and applicability of the rule clauses. Through these operations, the decision adaptation rules of the large model are optimized and updated.

[0173] Step S1551: Evaluate the impact of the effect deviation index data corresponding to the identified rule clauses, assess the severity level and scope of the effect deviation, and determine the comprehensive impact level of the corresponding rule clauses based on the predefined comprehensive evaluation rule mapping table. Sort the rule clauses according to the comprehensive impact level, and prioritize the target rule clauses whose comprehensive impact level exceeds the preset level. Specifically, for each target rule clause, compare the corresponding demand and decision mapping relationship data in the historical version of the large model decision adaptation rules with the actual scenario feedback data, calculate the specific data items that do not match, record the specific data manifestation of the rule deviation and the cause obtained through correlation analysis.

[0174] Impact assessment is performed on the effect deviation index data corresponding to the identified rule clauses. The assessment includes the severity level of the effect deviation (e.g., minor, moderate, severe), determined by the magnitude of the deviation and its impact on the overall effect; and the scope of impact level (e.g., local, regional, global), determined by the spatial range of the deviation's impact and the number of management objectives involved. Based on a predefined comprehensive assessment rule mapping table, the severity level and scope of impact level are combined to determine the comprehensive impact level of the corresponding rule clause. The comprehensive assessment rule mapping table specifies the comprehensive impact levels corresponding to different combinations of severity and scope of impact levels. Rule clauses are sorted according to their comprehensive impact level, with priority given to target rule clauses whose comprehensive impact level exceeds the preset level, ensuring that important issues are addressed first.

[0175] For each target rule clause, compare the corresponding demand and decision mapping relationship data in the historical version of the large model decision adaptation rules with the actual scenario feedback data. Compare each data item individually, and calculate the specific data items that do not match, such as incorrect mapping relationships or unreasonable condition parameters. Record the specific data manifestations of rule deviations, such as the magnitude and direction of the deviation value; and the causes obtained through correlation analysis, such as data collection errors, unreasonable model assumptions, or changes in the external environment.

[0176] Step S1552: Based on the specific data representation and causes of rule deviations, modify the data items in the corresponding demand and decision mapping relationship data table, and adjust the decision trigger condition parameters, decision content pointing identifiers, or execution constraint data in the rule clauses.

[0177] Based on the specific data manifestations and causes of rule deviations, the corresponding data items in the demand-decision mapping table are modified. If the deviation is caused by unreasonable decision trigger condition parameters, the value range or threshold of the trigger condition parameters is adjusted; if the decision content points to an incorrect identifier, the identifier is corrected to ensure the correct correspondence between the decision content and the demand; if the execution constraint data does not match the actual situation, the execution constraint data is updated to conform to the actual constraints of the current scenario. By adjusting the above data items, errors and deviations in the rule clauses are corrected, improving the accuracy and applicability of the rule clauses.

[0178] Step S1553: Collect specific data on new changes and new requirements that are not covered, perform structured transformation on the collected data, and transform the description of the scene change features and the original text of the new requirements into structured decision requirement data entries, and record the decision target identifier and execution constraint parameters corresponding to the new requirement data.

[0179] Collect detailed data on newly emerging and uncovered change features and requirements, and then perform a structured transformation on this data. Convert unstructured descriptions of scenario changes and raw text of new requirements into structured decision-making requirement data entries, ensuring they conform to the data format requirements of the requirement-decision mapping table. During the structured transformation process, extract key information such as the type of change feature and the specific content of the new requirement. Record the decision-making objective identifier corresponding to the new requirement data, i.e., the management objective to which the new requirement belongs; and the execution constraint parameters, i.e., the constraints that must be followed to achieve the new requirement.

[0180] Step S1554: Based on the transformed decision requirement data items, generate a new rule clause data structure through the rule template. The new rule clause data structure contains the updated requirement-decision mapping relationship data, records the decision type code triggered by the new requirement, the decision content data, and the execution requirement parameters.

[0181] Based on the transformed decision-making requirement data entries, a new rule clause data structure is generated using a rule template. The rule template contains the standard format and necessary data items for the rule clauses. Information from the decision-making requirement data entries is populated into the rule template to generate the new rule clause data structure. This new rule clause data structure includes updated requirement-decision mapping data, clearly defining the correspondence between new requirements and corresponding decisions. It records the decision type code triggered by the new requirement, i.e., the decision type corresponding to that requirement; decision content data, i.e., the specific decision plan and operational suggestions; and execution requirement parameters, such as execution time, resource requirements, and quality standards. By generating this new rule clause data structure, the new change characteristics and new requirement characteristics are covered.

[0182] Step S1555: Add the newly generated rule clause data structure to the historical version large model decision adaptation rules, compare the logical relationship between the new rule clause data structure and the original rule clause data structure through the rule reasoning engine, update the content of the detected logical conflict content according to the preset conflict resolution rules, and establish a correlation index between the corresponding rule clauses for the identified logical complementary information.

[0183] The newly generated rule clause data structure is added to the historical version of the large model decision adaptation rules. The rule inference engine is invoked to compare and analyze the logical relationships between the new and existing rule clause data structures. Logical conflicts are detected, such as contradictory conditions or conflicting conclusions between rule clauses. For detected logical conflicts, content is updated and conflicts are eliminated according to pre-defined conflict resolution rules, such as priority rules, majority rules, and expert judgment rules. Simultaneously, logically complementary information between the new and existing rule clauses is identified; that is, they complement and improve each other in content. Association indexes are established between the corresponding rule clauses for this logically complementary information, facilitating the simultaneous consideration of related rule clauses during the decision-making process and improving the comprehensiveness and accuracy of the decision.

[0184] Step S1556: Use natural language processing tools to check the semantic consistency between the adjusted rule clauses and the newly added rule clauses, rewrite the ambiguous or unclear content based on the knowledge base, and adjust the logical connectors between the clauses.

[0185] Natural language processing tools were used to perform semantic consistency checks on the revised and newly added rule clauses. The semantic expression of the clause text was analyzed to ensure clarity and accuracy, and to identify any ambiguous, unclear, or inconsistent content. For problematic content, the clauses were rewritten based on standard terminology, definitions, and expressions from a knowledge base to make the semantics clearer and more consistent. Simultaneously, logical connectors between clauses, such as "and," "or," and "if...then...", were adjusted to ensure clear and coherent logical relationships between clauses. Through semantic consistency checks and text optimization, the readability and comprehensibility of the rule clauses were improved, avoiding decision-making errors caused by semantic issues.

[0186] Step S1557: Query the basic normative document library and historical decision-making case database of natural resource management, verify the compliance and rationality of the adjusted large model decision-making adaptation rules, and correct the identified clauses that do not meet the requirements of the basic norms.

[0187] The compliance and rationality of the adjusted large-scale model decision-making adaptation rules are verified by consulting the basic normative document library and historical decision-making case database for natural resource management. Compliance verification checks whether the rule clauses conform to the relevant norms and standards in the basic normative document library, ensuring the legality of the decision-making rules. Rationality verification refers to successful cases and experiences in the historical decision-making case database to assess the feasibility and effectiveness of the rule clauses in practical application. If clauses that do not meet the basic normative requirements are identified, they are revised to conform to the normative requirements; for unreasonable clauses, adjustments and optimizations are made based on historical case experience to improve the rationality and applicability of the rules.

[0188] Step S1558: Reorder the terms of the adjusted large model decision adaptation rules based on logical hierarchy, and output the adjusted large model decision adaptation rules.

[0189] All clauses of the adjusted large-scale model decision-making adaptation rules are reordered based on logical hierarchy. The clauses are organized into a hierarchical structure according to their logical relationships, such as general-to-specific, causal, and sequential relationships. Generally, general and principle-based clauses are placed first, followed by specific and detailed clauses; clauses with causal relationships are arranged in the order of cause first, then effect; and clauses with sequential relationships are arranged in the order of execution. This reordering makes the structure of the large-scale model decision-making adaptation rules clearer, the logic more rigorous, and easier to understand and apply. The final output is the adjusted large-scale model decision-making adaptation rules.

[0190] Step S156: Based on the adjusted large model decision adaptation rules and combined with the effect deviation index data, modify the specific management operation instruction sequence in the historical version management action combination scheme, replace the operation instructions whose effect index difference value exceeds the threshold, and adjust the time parameters in the execution sequence scheduling table of related management operations.

[0191] Based on the adjusted large-scale model decision adaptation rules and performance deviation index data, the historical version of the management action combination scheme was modified. Management operation instructions with performance index differences exceeding thresholds were identified, indicating that the execution effect of these instructions did not meet expectations. According to the adjusted decision rules and performance deviation analysis, these operation instructions were replaced, selecting more suitable operation types or adjusting operation parameters. Simultaneously, the time parameters in the execution sequence scheduling table of relevant management operations, such as start time, end time, and execution order, were adjusted to ensure coordination between operations and rational utilization of resources. By modifying the operation instruction sequence and timing scheduling table, the management action combination scheme was optimized, improving its execution effect.

[0192] Step S157: Retrieve management operation instructions that match the new scenario features from the management operation knowledge base, add them to the management operation instruction sequence of the management action combination scheme, and record the execution condition predicate, execution method code and execution timing parameters of the newly added management operation.

[0193] Based on the characteristics of the new scenario, matching management operation instructions are retrieved from the management operation knowledge base. New scenario characteristics may require new management operations or adjustments to existing operations. The retrieved matching management operation instructions are added to the management operation instruction sequence of the management action combination scheme to enhance the scheme's adaptability to new scenarios. Simultaneously, the execution condition predicates of the newly added management operations are recorded, i.e., the conditions that must be met for the operation to be executed; the execution method code, specifying the implementation method of the operation; and the execution timing parameters, such as the execution time window and duration. By adding new management operation instructions, the content of the management action combination scheme is enriched, improving its ability to cope with complex scenarios.

[0194] Step S158: Compare the optimized large model decision adaptation rules with the modified management action combination scheme, check the logical consistency between the rule adjustment content and the scheme modification content. If a conflict is found, adjust one of the rule adjustment content or the scheme modification content according to the preset conflict resolution strategy.

[0195] Compare the optimized large-scale model decision-making adaptation rules with the modified management action combination scheme to check their logical consistency. Analyze whether the rule adjustments and scheme modifications match and support each other, and whether there are any logical conflicts, such as rules requiring a certain operation that is not included in the scheme, or operations in the scheme contradicting constraints in the rules. If conflicts are found, handle them according to the preset conflict resolution strategy. The conflict resolution strategy may prioritize ensuring the authority of the rules and adjust the scheme modifications to comply with the rule requirements; or prioritize the feasibility of the scheme and adjust the rule adjustments to adapt to the scheme; or comprehensively consider the importance and impact of both to seek the optimal compromise. By resolving conflicts, ensure the logical consistency between the optimized decision rules and action schemes, and improve the reliability and effectiveness of the decision-making scheme.

[0196] Step S159: Combining the historical iteration optimization record database, simulate and deduce the adjusted large model decision adaptation rules and management action combination scheme, and output the optimized decision scheme data package containing the optimized large model decision adaptation rules and the optimized management action combination scheme.

[0197] By incorporating a historical iterative optimization record database, which stores information such as decision rules, action plans, and execution effects from previous iterations, a simulation environment is constructed using historical data to simulate the adjusted large-scale model decision adaptation rules and management action combination schemes. During the simulation, the execution process of the decision rules and action plans is simulated under different scenario conditions, evaluating their expected effects and potential problems. Based on lessons learned from the historical iterative optimization records, adjustments and optimizations are made to address issues identified during the simulation. Upon completion of the simulation, an optimized decision scheme data package is output, containing the optimized large-scale model decision adaptation rules and the optimized management action combination scheme. This data package can be used to guide subsequent natural resource management decision-making practices.

[0198] In one exemplary embodiment, a large-scale model-assisted decision-making system for natural resource information management is provided. This system can be a terminal, server, etc., and its internal structure diagram can be as follows: Figure 2 As shown, this large-scale model-assisted decision-making system for natural resource information management includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides the environment for the operation of the operating system and computer programs in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, near-field communication, or other technologies. When the computer program is executed by the processor, it implements a large-scale model-assisted decision-making method for natural resource information management. The display unit is used to generate a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the shell of a large model-assisted decision-making system for natural resource information management, or external keyboards, touchpads, or mice, etc.

[0199] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A large-scale model-assisted decision-making method applied to the information management of natural resources, characterized in that, The method includes: The system receives and identifies the core requirements input of a natural resource management scenario, and acquires the type feature data and real-time change data stream of natural resources. The core requirements input, type feature data and real-time change data stream are input into a pre-trained natural language processing model and knowledge graph for entity recognition, relation extraction and constraint extraction, generating scenario requirement analysis results. The scenario requirement analysis results include a management target direction data list and a scenario constraint data list. Based on the scenario requirement analysis results, large model decision adaptation rules are generated. The scenario requirement analysis results, the basic normative document library of natural resource management, and the historical decision case database are taken as input and processed by the rule building engine to generate large model decision adaptation rules. The large model decision adaptation rules drive the pre-trained natural resource management large model to output management action combination schemes. The large model decision adaptation rules and the current status data stream of natural resources obtained from the monitoring terminal are input into the decision generation module of the large model for processing to generate management action combination schemes. The management action combination scheme is issued to the execution unit in the natural resource management scenario. The data stream of resource status changes, operation execution effect logs and scenario environment change information during the scheme execution process are collected by the sensor network and execution log collection system deployed in the scenario. After cleaning, fusion and formatting, a multi-dimensional feedback dataset is generated. The large model decision adaptation rules and the management action combination scheme are iteratively optimized using the multi-dimensional feedback dataset. The multi-dimensional feedback dataset, the historical version of the large model decision adaptation rules, and the historical version of the management action combination scheme are input into the iterative optimization algorithm module for processing to generate an optimized decision scheme data package. The method involves driving the pre-trained natural resource management big model through the big model decision adaptation rules to output a management action combination scheme. This scheme involves inputting the big model decision adaptation rules and the current state data stream of natural resources obtained from the monitoring terminal into the decision generation module of the big model for processing, generating the management action combination scheme, including: Load a pre-trained large-scale natural resource management model, which includes a data preprocessing module, a rule parsing module, a decision generation module, and a result optimization module; The current state data stream of natural resources is input into the data preprocessing module of the large model for missing value imputation, outlier processing and standardization transformation to form a standardized state data matrix. The current state data stream of natural resources includes a real-time monitoring data set consisting of resource quantity time series data, resource distribution spatial data, resource quality index data and resource surrounding environmental parameter data. The large model decision adaptation rules are input into the rule parsing module of the large model. The mapping relationship data table between requirements and decisions and the list of execution constraints in the large model decision adaptation rules are parsed syntactically and semantically. The key decision element feature vectors and execution condition predicates in the rules are extracted to form the rule parsing result data structure. The standardized state data matrix and the rule parsing result data structure are input into the decision generation module of the large model. The decision generation module performs feature matching in the standardized state data matrix based on the key decision element feature vector in the rule parsing result data structure and searches in the preset management operation knowledge base. The management operation types with a matching degree higher than the key decision element feature vector are encoded and output as the filtering results. Based on the execution condition predicate in the rule parsing result data structure, sorting constraints are solved for the selected management operation type codes, mutual influence and synergy measures between different management operation type code combinations are calculated, and the execution sequence of various operation codes is arranged to form a preliminary operation sequence scheduling framework. Based on the preliminary operation timing scheduling framework, the detailed operation template corresponding to each management operation type code is retrieved from the management operation knowledge base. The corresponding parameter values ​​in the standardized state data matrix are filled into the corresponding variable positions of the detailed operation template to generate detailed operation plan data containing operation implementation object identifier, implementation method code, implementation scope coordinates and implementation resource list. The detailed operation plan data is input into the result optimization module of the large model. The result optimization module combines the operation effect index data in the historical decision case database, obtains the effect index of similar historical operations under similar parameters from the operation effect index database, compares it with the expected effect of the current operation parameters, updates and calculates the current operation parameters based on the comparison results, and generates detailed operation plan data after parameter adjustment. Analyze the resource consumption estimates for various operations in the detailed operation plan data, and combine them with the resource supply capacity data of the current natural resource management scenario to recalculate and allocate the resource allocation list in the operation plan; The adjusted detailed operation plan data undergoes a global constraint check to detect time windows that violate preset constraints in the operation execution timing schedule, and the time windows are recalculated and reassigned; logical conflicts between operation content data are identified and marked, and conflicting operations are selected or sorted according to preset priority rules; the demand and supply of each resource in the resource configuration list are checked, and data items with mismatched supply and demand are recalculated and reassigned to form a management action combination plan that includes a specific management operation instruction sequence and an execution timing schedule table; The step of iteratively optimizing the large model decision adaptation rules and the management action combination scheme using the multi-dimensional feedback dataset involves inputting the multi-dimensional feedback dataset, historical versions of the large model decision adaptation rules, and historical versions of the management action combination scheme into the iterative optimization algorithm module for processing, generating an optimized decision scheme data package, including: Receive a multi-dimensional feedback dataset, perform feature separation on the set of execution effect indicators and the set of scene change features in the multi-dimensional feedback dataset, split the set of execution effect indicators according to the effect dimension labels corresponding to the data list of management target directions, and divide the set of scene change features according to the change type label and the scope of impact label to form a subset of categorized feedback data; Compare the execution performance metrics data in the subset of categorized feedback data with the expected performance metrics data of the historical version management action combination scheme, calculate the difference value vector, identify the operation records where the performance metrics difference value exceeds the difference threshold, and record the operation identifier and performance deviation metrics data of the corresponding management operation. Based on the performance deviation index data, trace the rule clauses related to the corresponding operation identifiers in the decision adaptation rules of the historical version of the large model, compare the demand and decision mapping relationship data in the rule clauses with the current scenario feedback data, calculate the adaptation difference value, and locate the rule clause identifiers that cause performance deviation. By comparing the scenario change feature data in the classification feedback data subset with the rule application scope definition data of the historical version of the large model decision adaptation rules, we can identify the new change features and new demand features that appear in the current scenario, and record the specific data of the new change features and new demand features that are not covered by the historical version of the large model decision adaptation rules. Based on the identified rule clause identifiers and specific data on new changes and new requirements that are not covered, new requirement entries and decision mapping relationships corresponding to the new changes and new requirements are created in the historical version of the large model decision adaptation rules, and these are added to the mapping relationship data table as new rule clauses; rule clauses with deviations are identified, and decision mapping relationships or condition parameters that do not match the actual scenario feedback data are updated with the corrected content. Based on the adjusted large model decision adaptation rules and combined with the effect deviation index data, the specific management operation instruction sequence in the historical version management action combination scheme is modified, the operation instruction with the effect index difference value exceeding the threshold is replaced, and the time parameters in the execution sequence scheduling table of related management operations are adjusted. Retrieve management operation instructions that match the characteristics of the new scenario from the management operation knowledge base, add them to the management operation instruction sequence of the management action combination scheme, and record the execution condition predicate, execution method code and execution timing parameters of the newly added management operation; Compare the optimized large model decision adaptation rules with the modified management action combination scheme, check the logical consistency between the rule adjustment content and the scheme modification content, and if a conflict is found, adjust one of the rule adjustment content or the scheme modification content according to the preset conflict resolution strategy. By combining historical iteration optimization record databases, the adjusted large model decision adaptation rules and management action combination schemes are simulated and deduced, and the optimized decision scheme data package containing the optimized large model decision adaptation rules and optimized management action combination schemes is output.

2. The large-scale model-assisted decision-making method for natural resource information management according to claim 1, characterized in that, The process of generating large-scale model decision adaptation rules based on the scenario requirement analysis results involves taking the scenario requirement analysis results, the basic normative document library for natural resource management, and the historical decision case database as inputs, processing them through a rule building engine, and generating large-scale model decision adaptation rules, including: The management target direction data list in the scenario requirement analysis result is semantically segmented and clustered into multiple specific target item data, and a performance-oriented label predefined by the system is attached to each specific target item data. At the same time, the scenario constraint data list is extracted from the scenario requirement analysis result, and each constraint data is labeled with an applicable scenario scope label predefined by the system and an impact dimension label predefined by the system. The basic normative document library for natural resource management is retrieved. The basic normative document library for natural resource management includes a collection of documents consisting of resource protection norms, development and utilization norms, ecological maintenance norms, and supervision and management norms. The text matching algorithm is used to scan the clauses in the basic normative document library one by one, calculate the correlation score between each clause and each data in the current scenario requirement analysis result, and filter out the normative clauses with a correlation score higher than a preset threshold. The historical decision case database is obtained, and cluster analysis is performed on the records in the historical decision case database. The historical decision case records are automatically classified and indexed according to the tag system of the management target direction data list and the scenario constraint data list. The historical decision case database contains historical decision instance records with similar characteristics to the current management scenario. Each historical decision case record contains scenario requirement information field, decision content field, execution process log field and final effect index field. The specific target data after being split is associated and mapped with the filtered normative clauses. Based on the performance-oriented labels of the specific target data, it is bound to the normative clauses that meet its constraints. By traversing the mapping operation, the mapping relationship between each specific target data and at least one normative clause is established, and a mapping relationship table is generated. Pattern mining is performed on the classified historical decision-making case records to extract the decision logic graph and execution critical path of each type of historical decision-making case. The association patterns between the scenario requirement information field and the decision content field in the historical decision-making case are analyzed. The adjustment patterns of the decision content field under different scenario constraints are summarized, and the summarized patterns are stored as a historical decision-making case decision pattern knowledge base. Based on the mapping relationship table and the historical decision case decision pattern knowledge base, an initial decision rule set is constructed by filling and combining rule templates. The initial decision rule set includes rule structure element data, relationship data of each element, and rule application scope definition data. The initial set of decision rules is matched and merged with the list of scenario constraint data in the scenario requirement analysis results. The execution constraint data in the rule set is modified to make the constraint data in the rule set consistent with the current scenario constraint data list, thus forming a preliminary set of decision adaptation rules. The initial decision-making adaptation rule set is logically consistent. The logical relationships between the clauses within the rules are compared. Clauses with logical conflicts are marked as pending. Based on the preset conflict resolution rules or by calling the knowledge base, reasoning is performed to generate replacement clauses to update the original conflicting clauses. For clauses with missing information, supplementary content is generated based on the relevant information in the knowledge base to obtain a complete initial decision-making adaptation rule set. The improved preliminary decision adaptation rule set is verified against the basic normative document library of natural resource management. The differences between the rule clauses and the core requirements of the basic norms are compared. The inconsistent clauses are updated according to the core requirements of the basic norms. The large model decision adaptation rules, which include a data table of the mapping relationship between demand and decision and a list of execution constraints, are output.

3. The large-scale model-assisted decision-making method for natural resource information management according to claim 2, characterized in that, The clustering analysis operation on the records in the historical decision-making case database, which automatically classifies and indexes the historical decision-making case records according to the label system of the management objective direction data list and the scenario constraint condition data list, includes: Extract the scenario requirement information field from each record in the historical decision case database, and parse the management objective description text and scenario restriction description text for each historical decision case from the scenario requirement information field; Calculate the semantic similarity between the management objective description text of each historical decision case and the management objective direction data list in the scenario requirement analysis results. According to the preset similarity threshold, the historical decision cases are initially divided into core objective matching case set, partial objective matching case set and objective mismatch case set, and the objective mismatch case set is removed from the current processing queue. Extract key restriction feature words from the scenario restriction description text of each case in the initially divided core target matching case set and partial target matching case set, and perform feature matching calculations with the scenario restriction condition data list in the scenario requirement analysis results; Based on the matching calculation results of key restriction feature words, the core target matching case set is further divided into a fully restricted matching case subset and a partially restricted matching case subset. The partially target matching case set is divided into a partially target matching case subset whose key restriction feature matching number reaches a preset threshold and a partially target matching case subset whose key restriction feature matching number does not reach a preset threshold. Extract the data features of each subset of cases after division, record the common decision-making logic pattern, typical execution path sequence, common operation type set and core performance index data of each subset of cases, and form a category feature description file; Based on the category feature description file, each subset of historical decision cases is assigned a system-generated category identifier and a feature-generated applicable scenario description text. Historical decision cases within the same category are sorted by time according to decision timestamps. The adjustment trends of decision parameters in historical decision cases at different time periods are analyzed, and the influence pattern data of time factors on decision parameters are summarized. Extract typical historical decision-making cases from each subset of historical decision-making cases whose decision-making performance indicators exceed a preset threshold, and record the unique identifier and key decision-making step sequence of the typical historical decision-making cases. The missing key information fields in the records of each category of historical decision-making case subsets are inferred and completed based on association rules. Based on the completed decision-making process data, execution detail data, and effect feedback data of each category of historical decision-making case records, the case classification index is formed by integrating them according to the classification labels.

4. The large-scale model-assisted decision-making method for natural resource information management according to claim 1, characterized in that, The process involves solving the sorting constraints on the selected management operation type codes based on the execution condition predicates in the rule parsing result data structure, calculating the mutual influence and synergy measures between different combinations of management operation type codes, arranging the execution sequence of various operation codes, and forming a preliminary operation sequence scheduling framework, including: Extract the list of execution condition predicates from the data structure of the rule parsing results, and parse out the execution prerequisite predicates, execution time interval predicates, and execution order predicates for various management operations. Organize all the parsed execution condition predicates into a list of constraint condition predicates. Extract the functional descriptions and attribute parameters corresponding to the selected management operation type codes, record the implementation cycle parameters, the list of resource types required for implementation, the scope of implementation impact, and the description of subsequent impacts after implementation for each type of management operation, and form a list of operation feature parameters; Based on the list of constraint predicates and the list of operation feature parameters, a directed graph construction algorithm is used to identify the dependencies between various management operations, mark management operations with subsequent dependencies, management operations that can be executed in parallel, and management operations with mutual exclusion, and generate a directed graph of operation dependencies. Calculate the management effectiveness prediction data corresponding to different management operation combination sequences, call the predefined synergy effect calculation function, calculate the synergy effect value of different management operation combinations, put the combinations with positive synergy effect values ​​first in the sorting sequence, and avoid management operation sequences that may produce negative mutual influence based on the predefined conflict detection model. Based on the overall target direction data list of natural resource management, and according to the contribution calculation model of management operations to target items, the execution priority weight of management operations is calculated. Operations with a contribution weight higher than a preset threshold are marked as priority execution level and assigned higher priority during sorting. Based on the directed graph of operation dependencies, synergy measurement data, and operation execution priority weights, the directed graph of operation dependencies, synergy measurement data, and operation execution priority weights are used as inputs. A time-series planning algorithm based on constraint satisfaction is called to perform calculations and output the execution sequence of various management operations. The planned execution stage code and planned time window of each management operation are recorded. By comparing the proposed execution sequence with the list of constraint predicates, the constraint satisfaction problem solver checks for violations of execution constraints and identifies issues such as resource supply conflicts or overlapping time windows. For the identified problems, the execution sequence of relevant management operations was adjusted, the start and end points of the planned time windows for each management operation were adjusted, the resource supply quota was recalculated and allocated, and the constraint satisfaction problem was re-executed for verification. Connect to a professional knowledge graph database in the field of natural resource management, use the entity relationships in the knowledge graph to evaluate and adjust the execution sequence, and adjust the management operation timing by referring to the standard process entities and practical case entity data in the professional knowledge graph. Based on the adjusted management operation execution sequence, planned execution phase coding, and planned time window, a preliminary operation timing scheduling framework is formed.

5. The large-scale model-assisted decision-making method for natural resource information management according to claim 1, characterized in that, Based on the identified rule clause identifiers and specific data on newly emerging and unaddressed features and requirements, the mapping relationship data table between requirements and decisions in the historical version of the large model decision adaptation rules is modified. New rule clauses are added to cover newly emerging scenario changes and requirements, and the content of rule clauses with deviations is corrected, including: The impact of the identified rule clauses is assessed by evaluating the severity and scope of the effect deviation. Based on a predefined comprehensive evaluation rule mapping table, the overall impact level of the corresponding rule clauses is determined. The rule clauses are then sorted according to their overall impact level, with priority given to those exceeding a preset level. For each target rule clause, the corresponding demand and decision mapping data in the historical version of the large model decision adaptation rules is compared with actual scenario feedback data. Specific data items showing mismatch are calculated, and the specific data manifestations of the rule deviation and the causes obtained through correlation analysis are recorded. Based on the specific data manifestations and causes of rule deviations, modify the data items in the corresponding demand and decision mapping data table, and adjust the decision trigger condition parameters, decision content pointers, or execution constraint data in the rule clauses. Collect specific data on new changes and new requirements that are not covered, perform structured transformation on the collected data, transform the description of scene changes and the original text of new requirements into structured decision requirement data entries, and record the decision target identifier and execution constraint parameters corresponding to the new requirement data; Based on the transformed decision requirement data items, a new rule clause data structure is generated through the rule template. The new rule clause data structure contains the updated requirement-decision mapping relationship data, and records the decision type code, decision content data and execution requirement parameters triggered by the new requirement. The newly generated rule clause data structure is added to the historical version of the large model decision adaptation rules. The logical relationship between the new rule clause data structure and the original rule clause data structure is compared through the rule reasoning engine. For the detected logical conflicts, the content is updated according to the preset conflict resolution rules. For the identified logical complementary information, the association index is established between the corresponding rule clauses. Use natural language processing tools to check the semantic consistency between the adjusted and newly added rule clauses, rewrite the ambiguous or unclear content based on the knowledge base, and adjust the logical connectors between clauses. Query the basic normative document library and historical decision-making case database of natural resource management to verify the compliance and rationality of the adjusted large model decision-making adaptation rules, and correct the identified clauses that do not meet the requirements of the basic norms. The terms of the adjusted large model decision adaptation rules are reordered based on logical hierarchy, and the adjusted large model decision adaptation rules are output.

6. The large-scale model-assisted decision-making method for natural resource information management according to claim 1, characterized in that, The management action combination scheme is then issued to the execution unit in the natural resource management scenario. After collecting resource status change data streams, operation execution effect logs, and scenario environment change information during the scheme execution process through a sensor network and execution log collection system deployed in the scenario, the data is cleaned, fused, and formatted to generate a multi-dimensional feedback dataset, including: Deploy a distributed data acquisition network that covers key areas and operation execution areas in the natural resource management scenario. The deployment coordinates and acquisition range parameters of the acquisition nodes are determined based on the operation implementation space range and resource distribution heat map data in the management action combination scheme. Configure the data acquisition node's acquisition frequency parameters and acquisition content identifier list. The acquisition frequency parameters are determined based on the management operation's execution cycle parameters and the resource status change rate model. The acquisition content identifier list includes resource status change information composed of natural resource quantity change time series data, quality change index data, and distribution change spatial data. Collect and manage the execution effect log data, including resource response index data after the operation is implemented, target achievement progress percentage data, resource consumption list during the operation implementation process, and operation execution smoothness index, which constitute a data stream reflecting the actual effect of the operation; Collect information streams on changes in the scene environment, including remote sensing monitoring data of the natural environment, text update streams of policy requirements, and parameter changes in technical application conditions, which constitute external environmental information affecting natural resource management. The collected data streams of resource status changes, operation execution effect logs, and scene environment changes are transmitted in real time, and the raw data collected is transmitted to the data processing center server through a preset data transmission protocol. The raw collected data transmitted to the data processing center server is classified and stored. According to the data pattern, the raw data is divided into resource status data table, execution effect data table and environmental change data table. Each type of data table is then mapped and stored according to the predefined data pattern. The data stored in categories is cleaned to obtain valid data records. Pre-defined standardization transformation rules are invoked to transform the field names, data formats, units of measurement, and codes of the valid data records to conform to the predefined standard data pattern, thus obtaining standardized data records. The standardized data records are then associated and integrated based on timestamps and spatial coordinates to form a multi-dimensional feedback dataset containing a set of execution effect indicators and a set of scene change features.

7. A large-scale model-assisted decision-making system for information management of natural resources, characterized in that, include: processor; A machine-readable storage medium for storing machine-executable instructions of the processor; The processor is configured to execute the large model-assisted decision-making method for natural resource information management as described in any one of claims 1 to 6 by executing the machine-executable instructions.

8. A computer program product, characterized in that, The computer program product includes machine-executable instructions stored in a computer-readable storage medium. The processor of the large-scale model-assisted decision-making system for natural resource information management reads the machine-executable instructions from the computer-readable storage medium and executes the machine-executable instructions, causing the large-scale model-assisted decision-making system for natural resource information management to perform the large-scale model-assisted decision-making method for natural resource information management as described in any one of claims 1 to 6.

Images