小波支持向量机在大气污染物浓度预测中的应用

用小波分解重构和支持向量机相结合的方法,建立大气污染物浓度预测模型。通过小波分解,将大气污染物浓度序列分解为不同频段的小波系数序列,再对各层的小波系数序列重构到原尺度上。利用相关分析的方法构建出低频小波系数a3和中频小波系数d3的支持向量机模型输入因子为前一天小波系数a3和7个气象因子;高频小波系数d2和d1以前三日的小波系数为输入因子,然后对各小波系数序列采用相应的支持向量机模型进行预测,各小波系数均使用ν-支持向量回归机(ν-SVR)算法和径向基函数,最后通过小波重构合成大气污染物浓度序列的最终预测结果。通过对大气SO2浓度预测实例证明,该大气污染物浓度预测模型具有推广能力较强、预测精度较高、训练速度快、便于建模等优点,具有良好的应用前景。

Application of Wavelet Support Vector Machine in Prediction of Air Pollutants Concentration

Based on method of wavelet and support vector machine (SVM),pollution concentration forecast model was established. Through wavelet transform concentration series was decomposed into coefficients subsequence on different scales,the coefficients subsequence was reconstructed. Input parameters of SVM model was selected by use of correlation analysis method. The input parameters of SVM model of low frequency coefficients and mille wavelet frequency coefficients included wavelet coefficients one day before and meteorological factors,while input parameters of SVM model of high frequency coefficients included wavelet coefficients three days before. The coefficients was predicted by using appropriate support vector machine respectively,finally after the synthesis of forecasting results of coefficients subsequences,final predicting result of air pollution concentration was obtained. Predicting results of SO2 concentration showed that the model based on wavelet and support vector machine exhibits its properties of high forecast accuracy,high generalization capability,fast training and easy modeling.