CN-122022185-A - Offshore ecological bearing capacity region division method based on artificial intelligence large model

CN122022185ACN 122022185 ACN122022185 ACN 122022185ACN-122022185-A

Abstract

The invention belongs to the technical field of ocean information, and relates to an offshore ecological bearing capacity region division method based on an artificial intelligence large model. The method comprises the following steps of multi-source data acquisition and preprocessing, constructing preprocessed data into a feature library of a dynamic ecological health condition evaluation index system, designating ecological bearing capacity tags as scene features, constructing a pressure-state-response dynamic ecological health condition evaluation index system, evaluating the ecological health condition of an object of interest, analyzing state parameters according to the state parameters in the state of the evaluation result, carrying out clustering analysis on the state ecological parameters to obtain a preliminary regional division result, and optimizing the obtained preliminary regional division result by adopting a multi-objective function to obtain a final regional division result. The method realizes accurate and real-time regional division through multi-source data fusion, AI large model dynamic evaluation and self-adaptive optimization division, and provides scientific decision support for ocean management.

Inventors

HU ZIYUAN
SUN XIAOXIA
SUN SONG

Assignees

中国科学院海洋研究所

Dates

Publication Date: 20260512
Application Date: 20260213

Claims (4)

1. The offshore ecological bearing capacity region division method based on the artificial intelligence large model is characterized by comprising the following steps of: The method comprises the steps of firstly, collecting and preprocessing multi-source data, wherein the multi-source data comprises structured data and unstructured data, the structured data comprises marine organism diversity data, hydrological data, marine topography and topography data and human activity data, the unstructured data comprises ecological investigation reports and scientific literature, and the collected multi-source data is integrated and preprocessed; Step two, designating an ecological bearing capacity label as scene characteristics, constructing a dynamic ecological health condition evaluation index system based on a pressure-state-response model, analyzing the ecological health condition of the object of interest by adopting an analytic hierarchy process, and analyzing according to state parameters in the state of an evaluation result, wherein the ecological parameters in the state are used as the basis of the next analysis; step three, carrying out cluster analysis on the ecological parameters in the state of the step two to obtain a preliminary region division result; optimizing the obtained preliminary region division result by adopting a multi-objective function to obtain a final region division result, wherein the method specifically comprises the following steps of: (1) Maximization of ecological protection value ; Wherein, the Normalizing to [0,1] for the number of species within the regional grid cell; ; The habitat is divided into endangered species habitat, extremely endangered species habitat and other habitats; A ecological load-bearing index; Is a spatial weight; (2) Minimization of socioeconomic losses ; Wherein, the fishery is lost Travel heat loss Shipping loss ; The value range of (1, 0); (3) Management cost minimization ; Wherein the cost is monitored Law enforcement costs Ecological restoration cost ; The value range of (1, 0); (4)Minimize wherein The divided region numbers are referred to as decision variables.
2. The method for partitioning an offshore ecological bearing capacity region based on an artificial intelligence large model according to claim 1, wherein in the first step, the preprocessing comprises the steps of performing data cleaning, data normalization and data enhancement on unstructured data, and extracting characteristics related to the offshore ecological health condition; the pretreatment of unstructured data adopts an AI large model based on a transducer architecture to carry out named entity identification and relation extraction on text data, and an ecological function embedded vector is generated.
3. The method for partitioning an offshore ecological bearing capacity region based on an artificial intelligence large model according to claim 2, wherein the preprocessing of unstructured data comprises the following steps: (1) Inputting text data; (2) Performing entity-relationship triplet extraction by using a fine-tuning BERT model; (3) And inputting the triplet into an ecological knowledge graph embedding model to generate a high-dimensional vector.
4. The method for partitioning an offshore ecological bearing capacity region based on an artificial intelligence large model according to claim 3, wherein in the third step, UMAP is used for dimension reduction, and then K-means cluster analysis is performed.

Description

Offshore ecological bearing capacity region division method based on artificial intelligence large model Technical Field The invention belongs to the technical field of ocean information, and relates to an offshore ecological bearing capacity region division method based on an artificial intelligence large model. Background The traditional offshore ecological bearing capacity evaluation method is mostly dependent on static indexes and manual analysis, such as a analytic hierarchy process, an ecological footprint process, a fuzzy comprehensive evaluation method and the like, and has the problems of lag data update, strong subjectivity, incapability of dynamically adapting to complex environmental changes and the like. For example, the analytic hierarchy process is greatly influenced by expert subjective, the ecological footprint process has single evaluation factor, human-environment interaction is ignored, the membership function of the fuzzy comprehensive evaluation process is difficult to determine, and the calculation is complex. In addition, the existing method is difficult to process multi-source heterogeneous data (such as remote sensing, GIS, text and the like), and the regional division result lacks real-time property and self-adaptability. Therefore, it is necessary to design an offshore ecological bearing capacity region division method integrating artificial intelligence large models, dynamic clustering and multi-objective optimization to solve the above-mentioned drawbacks. Disclosure of Invention The method for dividing the offshore ecological bearing capacity region is automatic, intelligent and high in interpretability. In order to achieve the purpose, the technical scheme adopted by the invention is that the offshore ecological bearing capacity region dividing method based on the artificial intelligence large model comprises the following steps of: The method comprises the steps of firstly, collecting and preprocessing multi-source data, wherein the multi-source data comprises structured data and unstructured data, the structured data comprises marine organism diversity data, hydrological data, marine topography and topography data and human activity data, the unstructured data comprises ecological investigation reports and scientific literature, and the collected multi-source data is integrated and preprocessed; step two, designating an ecological bearing capacity label as scene characteristics, constructing a dynamic ecological health condition evaluation index system based on a pressure-state-response model, analyzing the system by adopting an analytic hierarchy process, and analyzing according to state parameters in an evaluation result state, wherein the ecological parameters in the state are used as the basis of the next analysis; step three, carrying out cluster analysis on the ecological parameters in the state of the step two to obtain a preliminary region division result; optimizing the obtained preliminary region division result by adopting a multi-objective function to obtain a final region division result, wherein the method specifically comprises the following steps of: (1) Maximization of ecological protection value ; Wherein, the Normalizing to [0,1] for the number of species within the regional grid cell;; is a habitat scarcity coefficient; A ecological load-bearing index; Is space weighted The habitat is divided into endangered species habitats (1.2), extremely endangered species habitats (1.5) and other habitats (1.0) as a scarce coefficient of the habitat; A physiological load-bearing index (0-1 normalized); for the spatial weight, can set; (2) Minimization of socioeconomic losses ; Wherein, the fishery is lostTravel heat lossShipping loss;The value range of (1, 0); (3) Management cost minimization ; Wherein the cost is monitoredLaw enforcement costsEcological restoration cost; The value range of (1, 0); (4)Minimize Wherein The divided grid numbers are referred to as decision variables. In the first step, the preprocessing comprises the steps of data cleaning, data normalization and data enhancement on unstructured data, extracting characteristics related to offshore ecological health conditions, and performing named entity recognition and relation extraction on text data by adopting an AI large model based on a transform architecture to generate an ecological function embedding vector for the unstructured data. Further, the preprocessing of unstructured data includes: (1) Inputting text data; (2) Performing entity-relationship triplet extraction by using a fine-tuning BERT model; (3) And inputting the triplet into an ecological knowledge graph embedding model to generate a high-dimensional vector. In the third step, the dimension is reduced by UMAP, and then K-means cluster analysis is performed. The method realizes accurate and real-time regional division through multi-source data fusion, AI large model dynamic evaluation and self-adaptive optimization division, and