围绕南方电网辖区范围内天生桥一级水电站(天一电站)的汛前(4-5月)径流的预测问题,引入年际增量预报概念,并与传统的预报模型进行比较。分析表明,径流的原值序列同时包含年际和年代际信号,预报难度较大,而年际增量序列则以年际信号为主。相关分析表明,增量序列放大了影响径流的因子信号,更易找到具有物理意义的气象因子。对预报模型进行十年的试预报检验,结果表明增量预报的丰枯一致率达到80%,并且抓住了2010-2013年的连续偏枯现象。预报相对误差合格率达到70%,同时很好地再现了天一电站汛前径流的1991-2005的上升趋势和2006-2013年的下降趋势。究其原因,是由于增量预报模型既较好地捕捉到径流增量的年际信号,同时又保留了实测径流的年代际信号。
This paper is focused on the long-term prediction of April-May runoff at Tianyi hydropower station,which is under the management of China Southern Power Grid,the inter-annual incremental approach has been applied to prediction model in this study.The original runoff series include both inter-annual and decadal signals,while the inter-annual runoff increment is mainly occupied by inter-annual signals.As a result,predicting the runoff increment may be much easier than predicting the original runoff.Based on the correlation analysis,more significant factors against the runoff increment than that of original runoff are chosen.The hindcast assessment shows that,by the incremental approach,the accordance rate of "flood or dry" is 80%,and it can catch the continuous drought phenomenon during 2010-2013.The pass rate of "relative error"" is 70%.Particularly,the incremental approach can reproduce the increasing trend during 1991-2005 and decreasing trend during 2006-2013.The possible reason for the good ability is the incremental approach can catch the inter-annual signal and keep the observed decadal signal.