Dynamic Behavior Analysis And Prediction Research On Ocean Environment Data

With the development of economy,ocean exploitation has gradually become a focus.Coastal countries are facing a series of environmental problems,such as coastal pollution and ocean disasters.The analysis and prediction of ocean environmental factors can effectively prevent the occurrence of pollution and disasters.They are key measures to maintain steady development of ocean ecological environment.Taking a certain sea area of our country as the specific research object,the prediction model of ocean environmental data is built based on information that is mined through cross recurrence plot and machine learning theory.Relevant findings have been applied in the public service platform deployed with ocean environment observation technology in Hebei ocean geological resources survey center.The main research contents are as follows.Firstly,for the non-stationary characteristics of ocean environmental data,a dynamic behavior analysis model is proposed.This model includes three functional parts: phase space reconstruction,recurrence matrix representation and cross recurrence plot visualization.Phase space reconstruction is performed to project data to high-dimensional space.Recurrence matrix represents the distance between trajectories in high-dimensional space.Cross recurrence plot is used to visualize the correlation between ocean data.Furthermore,a quantitative analysis mechanism is developed to quantify model performance.Simulation results show that this model can analyze the correlation between different ocean data from qualitative and quantitative perspectives.Secondly,aiming at the impacts of ocean environmental multi-factors,a multivariate prediction model is proposed.This model is consisted of analyzer and predictor.The factors that have strong impact on ocean disasters are selected autonomously by analyzer,which are served as the input of the following predictor.Here,the predictor is an ensemble model based on the stacking algorithm for multivariate prediction.Simulation results show that this model has lower prediction errors than other comparative models.Hence,it can accurately predict ocean environmental data.Finally,an ocean environmental monitoring platform is designed based on Matlab GUI.This platform consists of three functional modules,namely data storage,ocean environmental data analysis and prediction.The original ocean data and calculation results are stored in data storage module,while the correlation between ocean data is measured by analysis module.The prediction module is applied to multivariate ocean data prediction.Operation results show that this platform can achieve real-time display of ocean data series,analysis results and prediction results.Figure 26;Table 12;Reference 61...