溶解氧浓度的神经网络控制研究

污水生化处理过程常常受到入水流量水质变化而处于动态过程, 溶解氧浓度作为系统运行过程的一个关键变量, 采用经典的PI控制器难以保证良好的控制效果. 针对污水处理过程的溶解氧浓度控制问题, 提出了基于单神经元自适应PID算法和基于RBF神经网络两种控制器. 在国际基准Benchmark Simulation Model No.1 (BSM1)的仿真平台上进行仿真实验, 与经典PI控制器的运行结果对比, 表明了在所提出的两种控制器作用下, 溶解氧浓度具有更好的跟踪给定值能力, 控制系统具有更好的综合性能指标值.

Neural network control of dissolved oxygen concentration

The biological wastewater treatment process is always changing due to perturbations in influent conditions. Dissolved oxygen (DO) concentration, which is a key variable in the bioprocess, is difficult to be guaranteed good performance by the conventional PI controller. To solve the DO concentration control problem, neuron self adaptive PID controller and RBF neural network controller are presented. Then, the performance of the proposed controllers is verified by simulations on the platform of the Benchmark Simulation Model No.1 (BSM1). Compared with the PI controller, the proposed techniques applied for DO control have better tracking performance and the control systems have better overall performance index.