针对径流混沌时间序列固有的确定性和非线性特点,以传统的支持向量机预测模型为基础,通过引入模糊隶属函数来考虑气候和流域下垫面条件变化,建立了径流混沌时间序列的模糊支持向量机预测模型。通过三种方法对长江寸滩站实际月径流时间序列的预测模拟结果对比,表明本文建立的模型具有更高的模拟精度,是一种有效的径流时间序列预测模型。
This paper develops a prediction model of chaotic flow time series by using a fuzzy support vector machine based on traditional SVM and by analyzing the certainty and nonlinearity of these series.This model adopts a fuzzy membership function to quantify the effects of climate changes and the basin conditions.Three models of fuzzy SVM,traditional SVM and neural network are used to manipulate and predict the monthly flow time series at Cuntan of Yangtze River.A comparison shows that the fuzzy SVM is more accurate.