面向不平稳径流序列的中长期径流预测方法

摘要

Aiming at the short precision problem of long-term runoff forecast in unsteadily sequence,this paper built a forecast method which com-bined the trend analysis technology and the traditional periodic mean superposition technology. Firstly,the method separated the unsteadily runoff se-quence for trend sequences and the stationary series with stepwise regression trend analysis technology;secondly,to trend sequences,it used the trend expression extension to predict,and to the stationary series,it used periodic mean superposition technology to predict;finally,it got the final forcasting results combined the former two parts. The author took three stations of the middle Yellow River as an example,and the results show that using stepwise regression technology can be more exactly to express the trend of the sequence to improve the prediction accuracy,and it is better to combine trend analysis technology and periodic superposition technology to simulate the monthly runoff than using periodic superposition technology directly,which can achieve the long-term runoff forecast accuracy.%针对不平稳序列中长期径流预测精度低的问题，构建了趋势分析与周期叠加组合的预测方法。利用逐步回归趋势分析技术将不平稳径流序列分离为趋势序列和平稳序列两部分，对趋势序列采用趋势回归技术进行预测，对平稳序列采用周期均值叠加技术进行预测，然后将两者预测结果叠加得到最终的预测值。以黄河中游3个水文测站为例进行实例检验，结果表明：趋势序列采用逐步回归技术进行预测，能更准确表达序列的趋势性，从而提高预测精度；采用趋势分析与周期叠加组合的方法模拟月径流量，效果明显好于直接采用周期叠加的方法，能达到中长期径流预测的精度要求。