小波分析组合模型在日径流预测中的应用研究

针对日径流时间序列不断受人类活动干扰的影响，运用小波多分辨率功能，分别对黄河头道拐站和花园口站实测日径流时间序列进行小波分析，利用得到的低频成分和高频成分分别建立自回归模型，然后组合对日径流进行预测。与单一的原序列自回归模型相比，基于小波分析的组合模型预测精度有了显著提高，而且不同时期的预测精度基本一致：预测合格率都达到了90%以上。而单一自回归模型的预测精度在不同时期相差较大：人类活动影响越强，预测误差越大，尤其在花园口站的1969—1986 年和1987—2005 年两个时期的日径流预测合格率都小于85%。由此表明基于小波分析的组合模型对数据具有较强的抗干扰性，在径流预测方面有明显的优越性。

Application Research on Combined Models Based on Wavelet Analysis in Prediction of Daily Runoff

In the process of development and utilization of water resources, real-time and accurate daily runoff forecast plays an important role in the fields such as disaster reduction, flood control and other aspects. The process of the daily runoff time series is jammed increasingly by human activities. By the multi-resolution performance of wavelet analyis, the daily runoff time series are decomposed into the low frequency and the high frequency parts with which autogressive models are built seperately. These models are combined to predict separately the daily runoff time series of Toudaoguai station and Huayuankou station along the Yellow River in three periods of from 1966 to 1968, from 1969 to 1986 and from1987 to 2005. Compared with the single model gained by the original daily runoff time series, the prediction precision of the combined models based on wavelet analysis is increased obviously. The prediction precisions of three periods are consistent basically. Their prediction pass rates are more than 90% and predicative values can be used in practice. But the prediction precisions of the single model vary largely in three periods. With the aggravation of the human activities, the prediction errors increase accordingly. The prediction pass rates of Huayuankou station don't meet 85.0% in two periods from 1969 to 1986 and from 1987 to 2005. Their predicative results can only be referenced. So it is thought that the combined model based on wavelet analysis is more anti-jamming and has more significant superiority in runoff prediction.