Font Size: a A A

Study On River Flood Forecasting Based On LS - SVM

Posted on:2017-05-09Degree:MasterType:Thesis
Country:ChinaCandidate:H J DingFull Text:PDF
GTID:2132330488950179Subject:Control engineering
Abstract/Summary:PDF Full Text Request
Since China’s floods occur frequently, which will cause a very serious impact on our economy stable and healthy development. Flood forecasting as a very important non-engineering measures for flood control, it plays a huge role in the development of flood mitigation program of work process. This paper analyzed the relationship between the hydrological objects and water systems in the area of Fushun County, Zigong City, Sichuan Province, two methods of river flood forecast are used:water level forecasting and flow forecasting. In addition, because the impact of some external factors, some error data will be produced in the process of collecting the data of natural river water level and flow, we need to eliminate these error data to improve the prediction accuracy of water level and flow. In this paper, the following research work was done according to the above objectives:(1) This paper analyzed the reasons for how error data generated in the sample data, the classification of errors as well as the main features of the error data in hydrological tests, using the improved 3σ method and the chauvenet method in the error handling methods to deal with the error data, which is existing in the sample data.(2) This paper compared the least squares (LS), support vector machine (SVM), BP neural network and least squares support vector machine (LS-SVM) these four algorithms, it was concluded that the least squares support vector machines (LS-SVM)’s prediction ability is the best of them, so this paper used the least squares support vector machines (LS-SVM) to build prediction model.(3) By comparing the prediction effect of four kinds kernel function in the least squares support vector machines (LS-SVM), it was concluded that the RBF kernel function’s prediction effect is the best of them, so this paper used the RBF kernel function and grid search method to optimize model parameters.(4) Combining with the related water level and flow data in Fushun County, Zigong City, Sichuan Province to build three prediction models. Through studying and training on historical data in the sample data, we can obtain the output data, which is established by the three prediction models, using the "hydrological information and forecasting norms" (SL250-2000) as the basis to evaluate the prediction accuracy of the three models.According to the experimental results and the "hydrological information and forecasting norms" (SL250-2000) obtained:the prediction results of the three prediction models are all good, and the prediction accuracy of the double input single output (water level and flow-water level) prediction model is better than the single input single output (water level-water level) prediction model and the single input single output (flow-flow) prediction model. This result shows that:If the input factor increases, then the accuracy of the prediction models will increase too. This result can provide a reference for studying the changes of the water level and flow of the certain river.
Keywords/Search Tags:flood forecasting, LS-SVM, error handling, prediction model, prediction accuracy
PDF Full Text Request
Related items