Development of a Wavelet-Neural Network Based Visual Modeling Tool Towards Improving Efficiency of Interactive Runoff Forecasting
Accurate forecasting of river runoff poses great challenge on hydrologists. Many data-driven hydrological models have been developed to facilitate runoff forecasting due to their rapid development times and minimum information requirements. However, there is a strong need for a user-friendly tool through which the user could construct simulation models graphically instead of writing simulation code. To meet this challenge, a visual modeling tool that integrates visual modeling technology with a hybrid wavelet-neural network model has been developed with the goal of improving the modeling efficiency and enhancing the flexibility for interactive runoff forecasting. Within the tool, the wavelet transform is used to decompose the time series of runoff, then the sub-signals are used as input to the ANN model to predict the runoff for certain time step ahead.Based on an open modeling interface, the tool enables to connect different models in a single environment with visual feedback and real-time interaction to support simulation of a complex system of river dynamics, thus isolating users from the complexity of underlying programming. A case study for interactive daily runoff forecasting is given for Shule River basin located in northwest of China. A serial simulation experiments were carried out to verify the feasibility of the tool, and the results show that the tool can greatly enhance the easy-to-use capabilities of river runoff modeling.
forecasting wavelet artificial nerual network visual modeling runoff forecasting time series
Mutao Huang Yong Tian
College of Hydropower & Information Engineering Huzhong University of Science & Technology Wuhan Chi College of Hydropower & Information Engineering line Huzhong University of Science & Technology Wuha
国际会议
秦皇岛
英文
408-411
2010-11-05（万方平台首次上网日期，不代表论文的发表时间）