Research And Application Of Drainage Area Flood Forecast System Based On Multi-Agent

Agent and multi-agent technology is an effective way to solve complex system problems. In this thesis, Agent and multi-agent technology is used for the research of flood forecast system. The system structure, the flood forecast model based on BP-GA conjoint arithmetic and the agents communication and collaboration are the focus and difficulty of this thesis. The key technology of the development and implementation of flood forecast system based on multi-agent are studied to provide intelligent decision-making support for floodwater attemper, disaster assessment and assistant decision by managers, etc.The main research work involved in this thesis and the results obtained are as follows:(1) By analyzing the speciality and especial structure which the flood forecast system should possessed, the structure of flood forecast system based on multi-agent is established. Then, the main structure and structure of flood forecast node MAS are provided separately. It is abstracted the system into interface agent, management agent, information agent, flood forecast agent and result integration agent, and so on. Through the cooperation between each other, the identically goal of the system is accomplished.(2) Artificial neural networks and genetic algorithms are introduced to the flood forecast system, and at the same time, the forecast model based on BP-GA conjoint arithmetic is established. And the arithmetic is applied to the independent agent, thus combined the accuracy of neural network with the distribution of agent, which can improve system adaptability and the forecasting rationality effectively. By integrating the prediction results of all forecast agents, the problem of a single Agent which can not take account of various factors is sovled. Therefor, it is possible to exchange information in the internal structure of system, and that the accuracy of the prediction is improved.(3) By using java programming language combined with JADE development platform, the flood forecasting system based on multi-agent are designed. The main characteristic of multi-agent systems is fully used, the complex issues of decomposition, abstract and organizational functions. And the complex forecasting system would be disassembled into a number of mutual cooperation agent modules, which can be achieve good independent thinking. At the same time, using ACL agent communication language to process message correspondence, can improve the system with the ability of parallel processing problems. The functions of the system include: data management, watershed enquiries, attribute data management, flood forecasting, scheduling management and user management, and so on.(4) Take full advantage of flood data resource of Reihe basin, and use simulation platform for the simulation of the system. The simulation results show that the prediction error is lower than 6.09%, and the accuracy rate is over 92.4%. The results indicate that the system obtains higher accuracy of prediction than other counterpart systems.