Research On Key Methods Of Big Data Intelligent Analysis In Marine Environment

The ocean area occupies more than 70% of the earth and has huge natural resources and space resources.Therefore,ocean exploration is an important direction for the future development of human science.The reports of the 18 th and 19 th National Congress of the Communist Party of China all put forward major strategies for building a maritime power.The development of ocean exploration technology is a major task for safeguarding the sovereignty of the country’s territorial waters,developing marine resources,and safeguarding the national economy.In response to the needs of ocean development,this research combines big data and artificial intelligence technology to analyze ocean collected data in order to realize the utilization and value mining of massive ocean data.The main content includes four points:First,this dissertation establishes a marine environment multi-sensor data preprocessing algorithm model to relieve the pressure of uploading large-scale measurement data of marine survey ships,buoys and other equipment.The increasing number of sensors on the sea causes long information collection periods and large volumes,which brings inconvenience to the subsequent data uploading and processing work of the platform.This dissertation uses the "fog" computing technology of the Internet of Things to propose a data preprocessing model based on the fusion of outlier detection method,D-S evidence theory and DBN to achieve preliminary data quality control of marine sensor nodes and ensure the validity of uploaded data.Second,this dissertation establishes a multivariate analysis model of marine environmental data to solve the problem of marine data analysis with high dimensionality and high coupling characteristics.Different from other industry data,marine environmental information has characteristics such as temporal and spatial correlation and attribute correlation.When analyzing a single marine environmental element,it is necessary to understand the internal transformation rules of the entire marine environmental ecology in order to achieve accurate element analysis and prediction.This dissertation uses oceanic red tides as an example of analysis,using principal component analysis,clustering,association rules and other techniques to model the multi-dimensional environmental data such as sea temperature,salinity,inorganic salts,and dissolved oxygen that affect the growth and retreat of red tides,and establish a red tide environmental analysis model.This work realize the impact analysis of the complex environment inside the ocean red tide,and illustrate the role and effectiveness of machine learning algorithms in ocean applications.Third,this dissertation establishes a fuzzy time series forecasting model for marine environmental data to solve the classic time series forecasting problem in marine data.Different from the traditional time series forecasting of small-volume data sets,future marine environmental monitoring will face cross-airspace,large-volume,and time-efficient time-series analysis and prediction,and the main purpose of marine monitoring is to achieve changes in the future trend of the marine environment.Based on fuzzy theory and deep learning network method,this dissertation proposes a method for forecasting ocean fuzzy time series data,and uses the knowledge of ocean historical data to predict the future ocean environment elements in order to realize the practical value application of ocean big data.Fourth,based on the big data platform technology,this work builds a marine environment big data intelligent analysis system.The system is divided into data acquisition,data storage management,data processing,and data visualization multi-layer structure.The data crawler is used in the data acquisition module to attack NOAA,Argo and other public ocean data websites to download data sets.Spark technology is used for data storage and processing in the data management module.Web GL technology is applied in data visualization,and Echart and Three.js visualization libraries are cited.Meanwhile,a new contour calculation method has been established in this system to realize rapid surveying and mapping of sea chart contours with large volumes of data.On the whole,this system has realized the data analysis of ocean hydrology,ocean meteorology,ocean chemistry and other modules as a whole.In summary,the main purpose of this dissertation is to integrate artificial intelligence technology into the scientific work of the marine environment,and propose several marine environmental data processing models to solve the existing and future processing problems of large-scale ocean sensing data.For a long time,the simulation and prediction of the marine environment have mainly relied on the numerical models of physical equations.This work aims to use the advantages of cross-discipline to play the role of intelligent algorithms in marine data processing.And complement the advantages of intelligent technology and traditional ocean numerical models to realize the mining and utilization of ocean big data.