中文摘要：

人工湿地脱氮系统是一个多参数相互影响的复杂非线性系统,当采用传统单因素方法分析时无法比较各因素对脱氮效果的影响程度。为此,在中试研究的基础上,首次尝试采用遗传BP神经网络对湿地脱氮系统进行模拟,详细讨论了如何优选网络拓扑结构、训练样本规模和训练数据归一化方法等关键问题,建立了优化的遗传神经网络模型,并采用该模型进行了仿真计算。依据正交试验结果,确定了最佳运行工况,并将水位、水力停留时间等因素对脱氮效果的影响程度划分为四个等级,在此基础上有针对性地提出了可行的强化脱氮措施。

英文摘要：

Since nitrogen removal system of constructed wetland is a complex and nonlinear system affected by many interacting factors, the influencing degree of the factors on nitrogen removal effect is hard to be compared by traditional single-factor analysis methods. Based on pilot-scale experimental research, the genetic neural network was tentatively utilized for the first time to simulate the nitrogen removal system, and some key problems such as determining the optimal topological structure, sample scale and unitary method for training data, etc. were discussed. An optimized model for genetic neural network was established to conduct the simulative calculation. According to the results of orthogonal test, the optimal operation conditions were decided and the factors （e. g. water level, hydraulic retention time, etc. ） were classified into four grades. Also, proper feasible measures for enhancing nitrogen removal were put forward.