Research On Techniques Of Water Quality Time Series Data Processing And Early Warning Database Construction

In the recent years, with the development of the society and economy, accidents of water pollution occur frequently and the safety of drinking water becomes an urgent problem in our society. Analysis and prediction of the water quality, early or timely warnings of the emergent water pollution events are very important and necessary technical methods to guarantee the quality and safety of drinking water. Furthermore, these methods are also very effective to decrease the direct or indirect loss of pollution accidents. So, how to construct the drinking water's early warning system becomes a very urgent and significant project.Combined with the project on water quality early warning, the thesis mainly focuses on the processing of time series data and the establishment of early warning databases for water quality. It discusses the combinational algorithm of Rough Set (RS) theory and D-S evidence theory and its applications, as well as the design, actual construction and application of whole process of water quality early warning database.Aiming at the features of time series data and application requirements in early warning system, Rough Set (RS) theory can objectively process the uncertainty data and solve the discontinuity that may exist in the time series data. It reduces the conflicts and redundancy of data, and then concludes the rules of classification in order to predict the trends or grades of water quality. In the case of small amount and big fluctuation of data, the better prediction results can also be obtained.As we know, the domain size of the info-table greatly affects the results of the rules extracted. If the imformation is not sufficient enough, it is also difficlut to export enough rules. On the contrary, too many multi-attribute rules may lead to the conflict between the rules. In order to solve these problems, the thesis proposes the combinational algorithm of Rough Set (RS) theory and D-S evidence theory. With the evaluation based on objective knowledge using RS theory and analysis of subjective knowledge or empirical and uncertain information using D-S theory, mining of discipline on time series data of water quality and prediction on the trends or grades of water quality can be implemented efficiently. The experimental results indicate that this combinational algorithm can obtain better results than RS theory only when the data is missing, redundant or not enough. Generally, most of the databases mainly focus on the demand of management and maintain of its related data. According to the system architecture of the whole process of water quality early warning, the thesis completes requirements analysis, framework design, conceptual design, logical design and physical design. It proposes the rules of water quality early warning database, which will be a reference of the future research and actual construction. Based on this design, the actual construction techniques of the early warning database, which contains the building, optimization and also backup methods, are elaborated. Three level deployment scenarios are proposed, including the communication, data transaction between these levels, data flow of the whole process, and the up-layer application.The upper results have been applied in three cities'early warning systems, which are supplying information for the government departments and the waterworks.