提出采用递推算法多变量自回归模型进行各站年径流的生成，同时考虑径流的年、月特征，进行模糊最优聚类分析，据此生成各月径流量，从而建立了模糊随机生成模型．该模型既反映了各站之间的时空关系，又反映了月径流生成中径流总量与其年内分配之间的关系，比较符合径流的真实特征 A multi-variable autoregressive model based on recursive algorithm is proposed to generate annual runoff at each station. At the same time, the annual and monthly characteristics of runoff are taken into account, and fuzzy optimal clustering analysis is conducted to generate monthly runoff. Thus, a fuzzy random production model . The model not only reflects the spatio-temporal relationship between the stations, but also reflects the relationship between the total amount of runoff in the generation of monthly runoff and its distribution within the year, more in line with the true characteristics of runoff