Water Resources Planning and Management

Poster Abstract

Application of Wavelet Decomposition – Support Vector Machine Prediction Model in Reservoir Annual Runoff Prediction

Wednesday, January 4
10:00 AM - 5:00 PM
Location: ChuLv Hall

I


Abstract: Annual inflow prediction of reservoir has long been a critical and difficult problem in water resources system planning and management. In order to improve the precision and reliability of reservoir annual runoff prediction, a hybrid prediction model combined wavelet decomposition and Support Vector Machine is proposed. The model uses wavelet theory to deconstruct annual runoff with complicated information into several stationary series; predict each series separately by Support Vector Machine model and then reconstruct predicted results. Two prediction models are applied independently to predict annual runoff during 1995-2014 of Meishan reservoir, which is an important flood control reservoir of Huai River in China: 1. Support Vector Machine based inflow prediction model using actual inflows information; 2. Support Vector Machine based inflow prediction with three-scale wavelet decomposition to pre-process the inflow information. Result shows that annual average prediction error reaches 62% under Support Vector Machine model without wavelet decomposition treatment; while using three-scale wavelet decomposition, the average annual prediction error decreased to 9%, and the prediction qualified rate reaches 85%, meeting A-degree according to China’s hydrological prediction standard. Comprehensive analysis of the two models results shows that wavelet decomposition based pre-process treatment makes significant progress in improving the prediction performance, and Support Vector Machine based prediction method can better excavate annual runoff process feature even in the same case study. This paper provides a practical and effective method and idea for reservoir mid-long term runoff prediction problem.


Key words: annual runoff prediction; wavelet decomposition; Support Vector Machine; Meishan reservoir

Ting Zhou

lecturer
Anhui Agricultural University

Presentation(s):

Send Email for Ting Zhou


Assets

Application of Wavelet Decomposition – Support Vector Machine Prediction Model in Reservoir Annual Runoff Prediction



Attendees who have favorited this

Send Email for Application of Wavelet Decomposition – Support Vector Machine Prediction Model in Reservoir Annual Runoff Prediction