基于贝叶斯网络的梯级水库群漫坝风险分析（特约稿）

为克服现有大坝风险分析方法多针对单库大坝、且不能有效考虑不确定性因素对风险评估结果的影响这一研究局限，本文选取能有效处理不确定性问题的贝叶斯网络理论对梯级水库群展开大坝失效风险研究。基于统计资料，结合专家经验调查法，将超标洪水、上游溃坝洪水和强地震确定为水库漫坝的关键风险因素，建立了三种因素单独、组合作用下的单库、梯级系统漫坝贝叶斯网络风险分析模型，并将其应用到大渡河流域上下相连的猴子岩-长河坝两座干流梯级上。结果表明风险源单独、组合作用下的梯级水库群的漫坝风险的量级都较小，其中猴子岩的上游溃坝洪水所致的漫坝风险量级最小，这与其上游双江口水库为干流控制性梯级的特性有关。两库均被识别为薄弱梯级，对其漫坝风险起主导作用的风险因素分别为超标洪水和强地震，这将作为系统风险防控措施制定的重要考量。通过实例应用，验证了文中所提模型在梯级水库群风险分析中的有效性。求解过程快速、有效，对风险因子间的相关性、系统薄弱梯级的识别直观、清晰，根据结果能及时掌握风险源单独或组合作用下的梯级水库群风险变化并制定相应的风险决策，有利于后续风险管理环节的迅速开展，为水利水电工程风险分析领域的研究提供了新的研究手段。

Risk Analysis of Dam Overtopping for Cascade Reservoirs Based on Bayesian Network (Invited Contribution)

Existing dam risk analysis methods are suitable largely for a single reservoir/dam and the influence of uncertainty on risk assessment result is not taken into consideration.To overcome these limitations,the Bayesian network theory which can effectively deal with the uncertainty problem was adopted to study the risk of dam failure in a cascade reservoir group.In combination of statistics data and experts'' experiences,extreme flood,upstream dam-break flood and strong earthquake were identified as the key risk factors of dam overtopping. Bayesian risk analysis network models of dam overtopping for a single reservoir and successive dam breaking for two cascade reservoirs under single or compound effects of extreme flood,earthquake and upstream dam-break flood were established and used to analyze dam overtopping risk of Houziyan and Changheba,two successive cascade reservoirs in the Dadu river basin,southwestern China.Results show that the overtopping risk magnitudes of cascaded reservoirs under single and compound effects of risk sources are relatively low,and the magnitude of overtopping risk caused by upstream dam-break flood of Houziyan is the least,which is correlated with characteristics of the upstream Shuangjiangkou reservoir,a control cascade in the basin. Both Houziyan and Changheba reservoirs are identified as weak cascades,and their main leading risk factors are excessive flood and strong earthquake,which can serve as an important support for formulating systemic risk prevention and control measures. Validity and practicability of the proposed model were verified through the case study.The solving process was fast and effective,and the identification of the risk factors correlation and the weak cascade reservoir was intuitive and clear. Using model analyses,risk change of cascade reservoirs under single and compound effects of risk sources can be obtained in time and corresponding risk decisions can be made,which is beneficial to rapid development of follow-up risk management and provides a new research way for risk analysis research in hydraulic and hydropower engineering.