西江流域由于大量兴建堤防工程导致洪水归槽,改变了原天然河道的洪水槽蓄关系,使得用于防洪规划和洪水灾害风险评估的洪水序列失去了一致性。目前洪水归槽问题研究主要集中在出槽洪水的归槽还原计算方面,对于干支流洪水与区间洪水的遭遇考虑不足,对洪水发生变异的成因及非一致性洪水的频率计算方法也缺乏系统性的研究。本文提出基于小波分析的非一致性洪水频率计算方法,并在西江梧州站作了实例验证。该方法首先利用小波分析将非一致性洪水序列分解成确定性趋势成分和随机性成分;再采用非线性函数(如幂函数、对数函数、指数函数、多项式函数等)对趋势成分进行拟合,得到非一致性洪水序列在时间域上的确定性规律;采用P—III型频率曲线对随机性成分进行频率分布拟合,得到洪水序列在频率域上的随机性规律;采用分布合成法对确定性的预测值和随机性的设计值进行合成计算,再采用传统的一致性频率计算方法推求合成序列的频率分布,据此可以得到过去、现状、未来不同时期的洪水频率分布规律,从而可以推求变化环境下河道断面的洪水频率分布,为防洪规划和洪水灾害风险评估提供设计依据。
Massive embankments construction on the West River basin has leaded to flood returning to main channel not only, but caused significant changes also in flood channel storage relationship of the river, and hence the flood sequences for risk assessment of flood disaster after this project must be inconsistent with or different from those for flood control planning. The previous studies of flood returning to main channel focus on a recovery computation from the channel overflowing state back to the original one through modification. They lack a sufficient consideration of the encountering of mainstream and tributary floods and the interval flood, and a systematic approach to the causes of flood variation and the frequency analysis method of inconsistency flood. This article presents a frequency analysis method of such inconsistent floods based on wavelet analysis for calculation of flood frequency distribution in a river, with a case study of the Wuzhou station of the West River. First, by assumption of an inconsistent annual maximum discharge series, this method decomposes the flood series into a relatively consistent random component and an inconsistent deterministic component. Second, it fits the trend components with several nonlinear functions (power function, logarithmic function, index function, polynomial function, etc.) to obtain the consistency and certainty of the flood in time domain. Third, it determines the randomness of the flood sequence in frequency domain by fitting the frequency distribution of random component with a P-III frequency curve. Forth, it predicts the deterministic component and the design value of random components with a distribution synthesis technique, extracts the random component by the theory of decomposition and composition in time series analysis, and finishes the frequency analysis with a constrained and weighted fitting method. The last step is to synthesize the deterministic forecasting value and stochastic design value and thus obtain the past and present frequency dis