| 摘要: |
| 【目的】分析水文不确定性因素对径流预测的影响，提高中长期水文预报方法模拟预测结果的精度。【方法】将小波分析(WA)、人工神经网络(ANN)和随机分析联合使用建立径流预测模型，即在小波分析(WA)揭示流量时频特性的基础上，将径流原序列分为高频部分和低频部分，然后利用人工神经网络(ANN)对低频部分进行模拟预测，利用随机分析对高频部分进行分析，最后将各部分结果叠加作为最终预测结果。将所建立的径流预测模型用于渠江二级支流后河的径流预测，并与传统BP人工神经网络方法的预测结果进行对比。【结果】根据《水文情报预报规范》，以预测值的相对误差小于10%为标准，传统BP人工神经网络预测结果合格率为46.67%，而基于小波神经与随机分析的径流预测模型在正常水文年模拟预测结果的合格率为73.33%。【结论】基于小波与随机分析的径流模型预测精度好、合格率高，能得到更好的复杂水文条件下的径流预测值。 |
| 关键词: 径流预测 小波分析 人工神经网络 随机分析 |
| DOI： |
| 分类号: |
| 基金项目:美国能源基金会“中国可持续能源项目”（G 0610 08581） |
|
| Runoff forecasting based on wavelet analysis,artificial neural network and hydrologic frequency analysis |
|
LI Bao-qi,ZHOU Ze-jiang,MA Yan-bo
|
| Abstract: |
| 【Objective】This study aimed to improve the precision of runoff forecasting by analyzing the effects of hydrological uncertainties.【Method】A runoff forecasting model based on wavelet analysis (WA),artificial neural network (ANN),and hydrologic frequency analysis was constructed.First,the multi-time scale characters of hydrologic time series were analyzed using WA for understanding the internal structures of the series,then the high frequency part in original series was recognized by DWT and the rest was treated as low frequency series.Then ANN was used to simulate and forecast the low frequency series and the hydrologic frequency analysis method was used to predict high frequency series.Finally,the two parts were stacked as final forecasting results.The method was applied to the Hou River and compared with the results from BP and ANN analysis.【Result】The passing rate of ANN was 46.67% and that of the new model was 73.33% according to the Hydrology Information Specification with the relative error of less than 10%.【Conclusion】The WA,ANN and hydrologic frequency analysis based model had high precision and good qualified rate,and it could be used for complex hydrologic conditions. |
| Key words: runoff forecasting wavelet analysis artificial neural network hydrologic frequency analysis |
