针对堤防安全评价研究中影响因素选择与现行规范结合不紧密以及对堤防安全状态这个复杂系统的模糊性和随机性考虑不足的问题，以现行堤防安全评价相关规范为基础，构建由5个一级评价指标、22个二级评价指标、11个三级评价指标组成的层次指标体系。在此基础上，建立组合优化赋权模型对改进群组G1法和熵权法计算的权重进行组合赋权。通过引进云模型理论，把堤防安全评价的3种安全类别转化为对应的云评语，以浮动云和综合云计算得到堤防安全状态总目标云参数和云图，能直观确定堤防的安全类别，且与实际安全状态一致。同时，分析一级指标安全状态云图，锁定影响堤防安全的关键因素，对堤防安全状态发展的趋势作出分析，给出有针对性的建议。对于复杂堤防系统，基于组合赋权和云模型的方法更好地将模糊性和随机性结合起来，可为堤防安全综合评价和除险加固提供依据。

The water retaining structures built along rivers,canals,lakes,and coasts are called dikes,the earliest and widely used important flood control projects in the world.China has built more than 300,000 km of dikes above grade 5,which has played an important role in preventing floods and protecting people′s personal and property safety.Due to the problems of the low construction standard,poor quality and unstable foundation of some dikes,serious dangerous situations such as piping,landslide,bank collapse,and overflowing may even lead to dike breakin operation.Therefore,it is of great significance to evaluate the safety of the dike regularly and grasp the safety situation of the dike in time,to strengthen the management of the dike.Thus,dike safety evaluation was still mainly qualitative evaluation,lack of quantitative evaluation,so it is necessary to conduct more indepth research. Some studies only considered the influence of a single factor or much dependent on subjective experience of experts,which was not comprehensive and systematic,and there were few fuzzy and probabilistic studies considering the influencing factors of dike safety.Because dike safety is a multi-level,uncertain,and nonlinear dynamic system affected by many factors.The research on its evaluation method has always been a very complex problem.General numerical simulation,traditional standard evaluation and other methods were difficult to fully reflect the actual safety of dikes.The theory of cloud model evaluation method can comprehensively consider the fuzziness and probability characteristics of the evaluated object,and this theory has been widely used in all kinds of complex comprehensive evaluation research.Therefore,the cloud model theory was introduced into the dike safety evaluation.Based on the current relevant norms of dike safety evaluation,a hierarchical index system was constructed,which was composed of 5 first-level evaluation indexes,22 second-level evaluation indexes and 11 third-level evaluation indexes.Concerning the rating standards of the existing dike safety evaluation standards,based on the results of on-site quality inspection and safety review,the factors affecting dike safety were comprehensively considered through expert group decision-making.Each expert was given weight through the similarity of order relation to reflect the difference of different experts′ knowledge and experience.The traditional G1 method was improved,and the subjective weight of the index was calculated.The weight of the expert was calculated by entropy weight method to make the weight distribution of each index more scientific and reasonable.The corresponding three safety categories of dike safety were assigned to the corresponding scale interval and transformed into the corresponding safety cloud comments,and then the cloud comments of the third level evaluation index were determined by integrating expert opinions,and the floating cloud or integrated cloud computing model was used to fuse the cloud comments of different experts step by step.The total target cloud parameters of the dike safety state were obtained,and the corresponding cloud map was made,and the real safety state distribution of the dike was intuitively obtained.Based on this,the safety category was determined,and the development trend of the dike safety state was analyzed. Taking a dike on the left side of the Pearl River Channel as an example,the cloud evaluation of the third-level evaluation index was calculated using floating cloud,and the cloud parameters of the first-level index were obtained.The operation and management was rated as "unsafe" because of the great deficiency in the daily management and maintenance and safety monitoring of the dike,and the flood control capacity was rated as "unsafe" because the elevation of the dike was lower than the requirements of the design standards..Although the project quality was "basic safety",but the score is low,some "cloud droplets" fall into the "unsafe" scope.The main reason was that the dike took a long time to be completed and the construction standard was low.After many years of operation,the cross-section of the dike body,the clay content of the dike soil,the hidden danger of the dike body and the quality of slope protection have the problems of reduced safety,and the index value was on the low side.According to the cloud parameters of the first-level index and its corresponding combination optimization weight and overall goal of comprehensive cloud computing,the security state cloud map falls into the "basic security" range and was identified as the second-class dike,which was consistent with the conclusion of the traditional method of dike identification. The weight distribution of each index was optimized based on the scientific index system construction and combination weighting method,and the fuzziness and randomness of the influencing factors of dike safety were better combined by introducing cloud model theory.It is not only given a more accurate judgment of each state but also transforms the grade evaluation of dike safety into fuzzy and random comprehensive evaluation,compared with the traditional evaluation method.It can reflect the actual safety state of the dike more comprehensively,and provide a more scientific and reasonable basis for dike safety evaluation and strengthening management.