Font Size: a A A

Study On Computational Intelligence's Applications In Assessment And Forecasting Of Water Resources And Hydrological System

Posted on:2005-10-27Degree:DoctorType:Dissertation
Country:ChinaCandidate:Q G LiFull Text:PDF
GTID:1102360152975551Subject:Hydrology and water resources
Abstract/Summary:PDF Full Text Request
Hydrological and water resources system is a kind of complex and dynamic system, which have large amount of uncertain factor. Many problems of water resources and hydrological system always are complex, multi-dimension and nonlinear. Computing Intelligence (CI), which includes fuzzy sets theory, Artificial Neural Networks, Genetic and Algorithms etc., is more powerful in resolving these problems. Many fields of CI's application in assessment and forecasting of hydrological and water resources system are studied in this paper. The major contents and research results are as follows:Many assessments in water resources system such as post-assessment are multi-objectives and semi-structural problems that include qualitative objectives and quantitative objectives. A fuzzy multi-objective semi-structural analysis assessment method is proposes in this paper including the two above objectives and applied into post-assessment in water conservancy projects. Firstly, the qualitative objectives are determined by fuzzy group decision-making method. Then, a consolidation assessment method is given depending on the combination of the fuzzy recognition and fuzzy optimum model. Finally, a practical case shows that the methodology is reasonable and feasible.It's difficult to determine the weights of the indexes in water resources assessment. In order to resolve this problem and reduce the man-made influence and dynamic characteristic in assessment procedure, according to the fuzzy optimal neural network theory and model, the author suggests that the past assessment samples and standard class data of assessment have relation with class membership degrees respectively. The method is given to analysis and determined this relationship by neural network training. Finally, the assessment model for water resources assessment is established subsequently.Water resources classification is a complicated and nonlinear problem. Many kinds of neural networks have been used. Based on fuzzy set theory and Kohonen self-organization network, a new fuzzy clustering neural network is proposed in this paper. Integrated withART theory, the network has adaptive ability. Further, with the fuzzy competition learning algorithm, the network learning process is improved. A study case for regional water resources abundance assessment shows the feasibility of the model and methodology.This paper first introduces the support vector machine (SVM) regression forecasting method into hydrological forecasting. Further, based on the fuzzy recognition theory proposed by Prof. Chen Shouyu, a new kind of kernel function is proposed in the paper. The kernel function has a more reasonable physical significance. At the end, the results of a study case show that the SVM regression hydrological forecasting method and the fuzzy pattern recognition kernel function is reasonable and feasible.Areal rainfall is important basic data in a real time flood warning system. A good areal rainfall calculation means we can forecast flood more accurately and in time. Here, we propose an areal rainfall forecasting methodology integrated fuzzy optimum neural network with Geography Information System (GIS) methods. Using many models and methods provided by GIS software, we obtain more accurate areal rainfalls of a catchment. Then, these outputs of the GIS software are taken as the expected output of the fuzzy optimum neural network, and the network is trained to find the mapping between the areal rainfalls and observed rainfalls in all gauge stations. Finally, using the above mapping, new observed values are taken as input of the network, and we can obtain the catchment areal rainfall in time.The conclusions and the problems to be further studied are given at the end of the paper.
Keywords/Search Tags:water resources assessment, computing intelligence, fuzziness, fuzzy recognition, neural networks, fuzzy optimum neural network, intelligence forecasting, geographical information system, support vector machine
PDF Full Text Request
Related items