Neural Networks And Fuzzy Set Theory In Applied Research, Real-time Flood Forecasting

Flood is a useful resource as well as one of the natural disasters. In this case, how to reduce the loss and develop its benefit is an attractive research topic. Firstly, this paper discusses the characteristic of flood disasters and Chinese flood control conditions. Then, through analyzing the functions of engineering measures and non-engineering measures in flood control, the importance of Hydrological forecasting, one of the non-engineering, is expounded. Based on these, this paper discusses the development status of Hydrological forecasting model and the major research interests. The major contents and research results are listed as follows:1. Considering the complexity, uncertainty and nonlinear feature of Hydrological elements, this article applies the real-time flood forecast model of neural network with fuzzy pattern recognition (RFNF) for predicting real-time flood. The model combines the fuzzy mathematics with neural network, and presents a new method of Biggest Error Back Propagation (BE-BP) to train the model. With the case study of real-time flood forecasting, it proves that the forecast model is feasible and the BE-BP method is effective. It has more advantages than the single artificial neural network model. And the model is also an attempt to probe into a new method of Hydrological forecast.2. Because of the abbreviations and assumes of Hydrological forecast model and the complexity of Hydrological elements, such as, the protean of underlying surface elements and the uneven distribute of rainfall, the forecasted flow is different from the practical flow. This paper analyses the present methods of real-time correction and the condition of being generalized. Based on them, a method of fuzzy similar correction is given. The method integrates the similar of flow process, real-time of the error intonations and the experience of experts and forecasters. With the case study, it proves that the method of fuzzy similar correction is succinct and easy to use as well as valid.3. Making use of the modern technology of computer, the software for real-time flood forecast and correction is developed, which integrated with the RFNF and themethod of fuzzy similar correction. The software can be used to make conveniently alternation between the real-time flood forecast and correction by the visual and database technology. In this way, the value of them in the flood control is demonstrated adequately.Finally, a summary is given and some problems to be further studied are discussed.