小样本跳变水质时序数据预测方法

Forecasting method for water quality time series of few and abnormal data

根据三峡库区某些断面水质监测数据具有样本小、成库前后数据出现跳变的特点,提出一种适用于三峡库区的水质参数预测模型(ELS-SVM)。ELS-SVM通过建立数据预处理模型对原始小样本时序数据进行处理,增强了时序数据的平稳性,并使用模拟退火(SA)算法优选最小二乘支持向量机(LS-SVM)模型参数。与典型的小样本预测模型的比较实验表明,ELS-SVM模型更适用于三峡库区小样本水质时序数据的预测。

According to the characteristics of small samples and sudden change of some sections in Three Gorges, the ELS-SVM water quality prediction model for Three Gorges was proposed, which consisted of Equal Level Pretreatment Model （ELPM） and Least Squares Support Vector Machines （LS-SVM）. ELPM was brought forward to preprocess raw time series data to enhance the smoothness, and Simulated Annealing （SA） algorithm was used for choosing the optimal parameters of LS- SVM. The experimental results indicate that, compared to the typical forecasting models in limited samples prediction, the ELS-SVM forecasting model is superior to the others in water quality parameter prediction of Three Gorges.