神经网络PID控制器在湿法烟气脱硫中的应用

Application of Neural Network PID Controller to WFGD

针对在石灰石/石膏湿法烟气脱硫工艺中吸收塔浆液pH值变化过程存在高度非线性,时变性以及各种不确定性,常规PID控制难以达到满意的控制效果的问题,提出了一种基于Levenberg-Marquardt算法的神经网络PID控制器,对浆液pH值变化过程进行辨识和控制。该算法的本质是提供牛顿法的速度和保证收敛的最速下降法之间的一个折中,适合于性能指标是平方误差的神经网络训练。仿真实验表明,该算法收敛速度快,控制系统无超调,稳态精度高,能够快速跟踪浆液pH值的变化并对其进行有效控制,且具有较强的抗干扰能力,满足实时控制的要求。

To the problem that the change of the absorber slurry pH value is a nonlinear and time-variation process with a large number of uncertainties in the limestone-gypsum wet flue gas desulphurizafion technology, and difficult to meet the requirement of real-time control by conventional PID controllers, a neural network PID controller based on Levenberg-Marquardt algorithm is proposed to identify and control the change of pH value. This algorithm splits the difference between the Newton method and the steepest descent method, which is suitable for neural net- work training. The simulation results show that the system has properties of no overshoot, quick response and good steady accuracy, and the controller can quickly track and effectively control the change of pH with good anti-jamming ability, and satisfy the need of real-time control.