Intelligent Data Analysis And Application Research For Online Water Eclogical Environment Monitoring Of Three Gorges Reservoir Area

Three Gorges Project is by far the world's largest water conservancy and hydropower project,now it has been stepping into operational management period from project construction phase.To carry out a comprehensive follow-up work of the project,it is necessary to establish an Online Water Ecological Environment Monitoring System(OWEEMS)for Three Gorges Reservoir Area(TGRA),to provide services for sustainable development of ecosystem such as water ecological environment information perception,assistant decision,and emergency warning.With the rapid development of information technology,mankind has entered era of big data,capability of efficient and accurate processing for big data has become an important problem for many information system to solve.Principal Component Analysis(PCA)is a statistical method to grasp the principal contradiction of diverse aspects,it can analyze the main influencing factors from multiple things,simplify the solution of complex problems,and reveal the essence of things.How to combine PCA with other intelligent analysis techniques and quickly get valuable knowledge from massive data become a research focus in the field of data mining.At present,in the construction of OWEEMS for TGRA,it is still facing some problems of intelligent data analysis:poor forecasting accuracy of online monitoring time series data;lack of multi-granularity layer analysis method for water quality;lack of water quality evaluation method under the situation of incomplete information data;lack of accurate correlation analysis method to identify the relationship between water quality and its monitoring indicators,etc.In this dissertation,we combined researches to solve the problems based on PCA with the development of OWEEMS for TGRA,and propose some intelligent models or methods as follows:(1)A fused cloud model based on Calman filtering and time series analysis(FCMCT)is proposed.Aiming at problems of hardness to measure relationships among variables in high dimensional time series data of water ecological environment,and poor forecasting accuracy,etc.,a fused cloud model based on Calman filtering,Time Series Analysis model and Cloud Model is proposed,it improves the ability of analysis and prediction for time series water quality data,providing a powerful tool for analysis of water ecological condition in TGRA,experiments on monitoring data of TGRA demonstrate the effectiveness of the model.(2)A multi-granularity analysis model based on Fuzzy Time Series and PC A(MAFPCA)is proposed.Aiming at uncertainty problems such as randomness and fuzziness of time series data of water ecological environment monitoring for TGRA,and based on Fuzzy Time Series,PC A and Granular Computing,a multi-granularity analysis and prediction model is constructed with layers as:attributes layer,principal components layer and overall evaluation layer,joint calculation of Fuzzy Time Series analysis with variables of upper layer can be implemented at each level.Through the selection of multi-granularity hierarchy and the joint calculation,the model can reduce the noise effect of Fuzzy Time Series,improve prediction accuracy and robustness,so as to achieve accurate and efficient analysis of water quality changes at all levels.Experiments on data from TGRA and River Ganges GR2 monitoring station in Deep Bay Water Control Zone of northwest Hongkong show the model being effective and applicable.(3)A water quality evaluation model based on D-S synthetization of Neighborhood Rough Set and PCA(DSRPE)is proposed.Aiming at problems of too many indexes of water quality monitoring,high occurrence of data loss,low accuracy of water quality's evaluation and analysis,etc.,and combining the merits of Rough Sets and PCA in water quality evaluation,we propose a water quality evaluation model based on Neighborhood Rough Set,PCA and Dempster-Shafer Evidence Theory.Not only it can give water grades under the circumstance of incomplete information data,but also quantitative analyses,meeting the demand of water quality evaluation and analysis in different levels.Through experiments on UCI dataset and some water quality data of TGRA,results show it achieves to evaluate water quality in incomplete information systems,effectively reduces the amount of calculation for data analysis,and has good classification accuracy.(4)Methods to analyze the correlation between a factor and an object(CSFPCA)and analyze the correlation between different objects(COPCA)are proposed.There are often internal relationships between different objects in complex system,and the relationship between them is ususlly hidden deeply,analysis of correlation from external factors is an important development direction in the subject of association analysis,based on theory of PCA,we propose a method to evaluate the correlation between a possible factor and an object,and another method to evaluate the correlation between different objects,providing new ways to reveal correlations in big data area.Through experiments on UCI datasets and some water quality data of TGRA,results demonstrate their effectiveness,and reveal some seasonal variations of water quality and relationships between water quality indexes and eutrophication.