灰色关联度法与加权距离系数法在洪水相似性识别中的适用性分析
Analysis on the Applicability of Grey Relational Analysis and Weighted Distance Coefficient Methods in Flood Similarity Identification

以澴水花园水文站46场历史洪水为研究对象，筛选10个易量化、易获取的暴雨洪水特征指标，开展基于1013种指标组合(不包含单指标组合)的洪水相似性研究。建立以专家经验识别为基准、以匹配率和洪峰流量合格率为量化指标的评定标准，采用灰色关联度法和加权距离系数法进行相似洪水识别，并引入差分进化算法优化加权距离系数法权重。结果表明：降雨总量P、前期影响雨量Pa、洪峰流量Q、洪量W是识别相似洪水的关键指标；差分进化算法优化权重后，加权距离系数法识别效果显著提升；灰色关联度法识别结果更为稳健，对指标组合变化不敏感；仅采用事前指标的识别效果明显不如全过程指标。
Using 46 historical flood events recorded at the Huayuan Hydrological Station on the Huanshui River as the study object, this study selected ten quantifiable and readily available rainfall-flood indicators and conducted a flood similarity analysis based on 1013 indicator combinations (excluding single-indicator combinations). An evaluation criterion was established using expert experience-based identification as the benchmark, with the matching rate and the qualified rate of peak discharge as quantitative metrics. Similar floods were identified using the grey relational analysis (GRA) method and the weighted distance coefficient (WDC) method; the differential evolution (DE) algorithm was introduced to optimize the weights of the latter. The results show that total rainfall (P), antecedent precipitation index (Pa), peak discharge (Q), and flood volume (W) are key indicators for identifying similar floods. After the weights were optimized by the DE algorithm, the performance of the WDC method improved markedly. The GRA method produced more robust results than the WDC method and was less sensitive to variations in indicator combinations. Using only pre-event indicators led to markedly poorer identification performance than using full-process indicators.