洪水灾情评估实际上是一种模式识别问题，而人工神经网络模型具有逼近有界闭子集上任意非线性映射的特性，且模型更新方便．本文提出了基于遗传算法的洪水灾情评估神经网络模型，阐述了其基本原理和其法，实例研究表明其实用性、客观性和通用性． Flood disaster assessment is actually a pattern recognition problem, while artificial neural network model has the property of approximating any nonlinear mapping on bounded closed sub-sets, and the model updating is convenient. In this paper, a neural network model of flood disaster assessment based on genetic algorithm is presented, and its basic principle and its method are expounded. The case study shows its practicability, objectivity and universality.