Research On Wavelet Analysis's Application In Runoff Analysis And Forecast

Runoff is a natural phenomenon with tremendous uncertainty and complexity. In order to better understand its intrinsic mechanism and aid flood-control and power generation scheduling, researchers have brought about many approaches and undertaken in-depth studies. Much has been achieved. This thesis studies the long-term runoff analysis and forecast by means of wavelet analysis.At the beginning, the thesis briefs the general approaches in long-term runoff analysis and forecast as well as their advantages and shortcomings. Then the wavelet analysis theory is discussed. A comparison is conducted among wavelet analysis, Fourier Transform and Short-time Fourier Transform and wavelet analysis's strong ability in both time domain and frequency domain is highlighted. This chapter also deals with some generally used wavelets and algorithms.The thesis involves analyzing the future tendency and periodical characteristics of several runoff series, including yearly average runoff, year maximum runoff and monthly runoff. Some special conclusions drawn from the analysis can be used in runoff forecasting.Runoff forecast, which is the most important part of the thesis, is conducted. A wavelet-based AR model is introduced. Forecast results, which are compared with those achieved by common AR model, confirm that the introduction of wavelet analysis is conducive to higher forecast accuracy, although it could be further promoted.The thesis finally comments on the research and lists the future jobs.