结合神经网络的一般模型控制策略研究

Study on GMC combined with neural network control strategy

研究了一种结合神经网络的一般模型控制(GMC)策略,并将其应用于污水处理中有机碳去除过程的溶解氧(Dissolved Oxygen,DO)控制。利用神经网络建立有机碳去除过程的半动力学(灰箱)模型,模型中神经网络用于预测过程的耗氧率,其权值由克隆选择算法进行优化。仿真结果显示,经改进的GMC控制策略比传统的PI控制和经典的GMC控制有更高的控制精度。

In this paper, a GMC combined with neural network control strategy is proposed to control the dissolved oxygen in the organic carbon removal process of wastewater biological treatment. The neural network based model is used to represent the respiration rate in the process. The optimization work of the neural network is done by a clonal selection algorithm （CSA）. The simulation results show that the improved control strategy has better control accuracy compared to conventional PI control and classic GMC control.