基于递归高阶神经网络的污水处理系统建模

Wastewater Treatment System Modeling Based on High-Order Recurrent Neural Network

针对污水处理过程的多变量、非线性、大滞后和强耦合的特点,利用递归高阶神经网络对污水处理过程关键水质参数——化学需氧量、生化需氧量、悬浮固体和氨氮——进行建模.对污水处理厂生化反应过程实际运行数据的实验表明所提出的建模方法是有效的,同时与前馈神经网络建模和一阶递归神经网络建模相比较,结果显示递归高阶神经网络建模具有更高的精确性.

Aiming at the characteristics of wastewater treatment process such as multi-variable,nonlinearity,large time delay and strong coupling,a modeling method using recurrent high-order neural network（RHONN） is proposed to establish models for the key parameters,including chemical oxygen demand,biological oxygen demand,suspended solid and NH_4-N. This modeling method is then applied to a certain wastewater treatment plant＇s actual running data during biological reaction process.The simulation results demonstrate that the proposed method is effective.The comparisons with the feed-forward neural network and first-order recurrent neural network show mat the modeling results by the recurrent high-order neural network own higher accuracy.