基于预测模型的SNN控制在湿法脱硫中的应用

Application of SNN Control Based on Predictive Model in Wet Desulfurization

针对目前燃煤电厂湿法脱硫动态特性具有大迟延、大惯性、时变、非线性等特点,且系统自动控制投运率低的现状,该文设计了一种基于预测模型的单节点神经网络(SNN)控制算法,将单节点神经网络结构控制器与传统的PID控制规律相融合,既具有了自学习和自适应的特点,也弥补了常规PID参数不能在线调整带来的弊端。同时加入预测控制,将系统未来的输出值提前反馈到控制器,对过程纯滞后特性具有明显的补偿效果,提高系统的稳定性和鲁棒性。结合某300 MW CFB机组炉外湿法脱硫系统数学模型进行控制仿真。结果表明,该控制算法相对于传统PID控制不仅超调量小、调节时间短、而且能有效解决模型参数改变和扰动带来的不稳定性,具有较强的适应性和抗干扰能力。

Against the characteristics of the wet method of coal desulfurization,such as large delay,large inertia,timevarying and nonlinear characteristics,and the system of low rate of automatic control of commissioning. Single node neural network（SNN） control based on predictive model was designed. The single node structure of neural network controller with the traditional PID control law of integration,not only has the characteristics of self-learning and adaptive,but also make up for the conventional PID parameters cannot bring the disadvantages of online adjustment parameters. At the same time,the predictive control is added,the output value of the system is fed back to the controller in advance,which has obvious compensation effect on the process of the pure lag,and improves the stability and robustness of the system. Combined with the mathematical model of the wet desulfurization system for a CFB 300MW unit,the results show that the control algorithm is compared with the traditional PID control not only small overshoot,short adjusting time,and can effectively solve the model parameters and disturbance instability,with strong adaptability and anti-jamming ability.