中文摘要：

区域各异的水安全评价指标体系限制了评价方法的通用性,故而至今还未形成统一的水安全评价方法。各水安全评价指标的单位、量纲并不完全相同,致使不同指标的同级标准值差异很大。本文在遵循各区域不同的指标体系基础上,通过对区域水安全评价指标体系的规范化处理,对传统的概率神经网络模型进行改进,提出适用于多区域、多项指标的具有通用性和普适性的水安全评价概率神经网络,消除了评价指标数目多少的限制和区域性评价指标各异性的限制,克服了传统概率神经网络的使用局限性。将规范化的概率神经网络模型应用到山东省水安全评价、全国五省水安全评价实例分析,研究结果表明：该模型的评价等级与其他多种方法的评价等级一致,不同的区域和不同数目的评价指标体系均可以直接用于评价,省去了复杂的编程计算工作,为评价模型的普适化、简单化提供了一条途径。

英文摘要：

No universal water security evaluation method is available today due to regional differences in evaluation indexes system. Differences exist in the unit and dimension of each water security index and a large difference appears between different indexes even on the same grade. This article presents a universal probabilistic neural network （PNN） for water safety assessment based on a reference value selection for all the indicators and an appropriate standardization transformation formula. This model can overcome the limitation of traditional PNN, and it adopted a standardized PNN in the water security assessment for Shandong province and other five provinces in China. Results show the same assessment results of normalized PNN with other methods. The new model, free of the restriction in the total number of evaluation indexes, provides a new approach to development of universal and simplified evaluation models.