| 摘要: |
| [目的]提出一种基于小波分析-支持向量机(WA-SVM)的水库溶解氧预测模型,以期提高水库溶解氧的预测精度.[方法]通过小波分解,将原始复杂的溶解氧浓度序列分解到不同的高频和低频层次,对每层得到的分解重构序列分别采用支持向量机回归方法进行预测后,合成原始序列的预测值,将该模型应用到于桥水库溶解氧浓度序列的预测中,并与单独支持向量机(SVM)回归方法预测结果进行比较.[结果]WA-SVM方法预测精度较SVM方法有较大提高.其平均绝对百分比误差和均方根误差分别为0.049 37和0.345 3,而SVM方法的分别为0.084 93和0.631 9.[结论]WA-SVM方法综合运用了小波分析的多分辨特性和支持向量机的非线性回归功能,能够较准确地预测水库溶解氧浓度. |
| 关键词: 小波分析 支持向量机 溶解氧 预测模型 |
| DOI： |
| 分类号: |
| 基金项目:国家自然科学基金 |
|
| Predicting on reservoir dissolved oxygen based on WA-SVM |
|
|
| Abstract: |
| 【Objective】 To enhance the precision of reservoir dissolved oxygen forecast,wavelet analysis and support vector machine (WA-SVM) forecast model was established. 【Method】 The original complex dissolved oxygen concentration series were decomposed to different layers through wavelet analysis,then each layer forecasted by means of SVM,and finally the forecasting results of the original time series obtained by composition.The model was applied to predict the dissolved oxygen series of Yuqiao reservoir,and compared with the results of SVM. 【Result】 The results indicated that the forecast accuracy of WA-SVM was higher than that of SVM.The mean absolute percentage error (MAPE) and the root of mean square error (RMSE) of WA-SVM method was 0.049 37 and 0.345 3 respectively,whereas MAPE and RMSE of SVM method were 0.084 93 and 0.631 9. 【Conclusion】 WA-SVM method can accurately predict reservoir dissolved oxygen concentration because of comprehensively using multi-distinguishable character of wavelet analysis and nonlinear regression function of support vector machine. |
| Key words: wavelet analysis support vector machine dissolved oxygen predicting model |
