负荷实时跟踪精细化氨法脱硫智能控制系统

传统的氨法脱硫控制系统存在延迟时间较长、无法实现实时跟踪负荷的局限性。针对该问题提出的Smith预估补偿装置，通过抵消系统中的纯滞后环节来提高控制系统的实时性。虽然该方法有效解决了长延时问题，但系统中PID参数调整采用的是试错法并依赖于调试操作经验，偶然性和因人而异导致系统波动较大。本文提出了BP（back propagation）神经网络的PID参数整定方法，该方法能实现对任意非线性函数的逼近，通过神经网络学习得到最佳的比例、微分、积分系数组合。运用该方法建模并进行长时过程控制仿真，结果验证了算法的可行性，其误差小，大幅提高了氨法脱硫系统的实时性和稳定性，实现了智能化精准控制效果。

Load Real-time Tracking Refined Ammonia Desulfurization Intelligent Control System

The traditional ammonia desulfurization control system has a long delay time and cannot realize real-time tracking of load. Aiming at this problem, Smith''s predictive compensation device is proposed to improve the stability of the control system by canceling the pure hysteresis in the system. Although this method effectively solves the problem of lag, the tuning of the PID parameters in the system still uses the trial and error method. This method is mainly adjusted by experience, which is very time consuming and has no clear judgment standard. A method for PID parameter tuning of BP (back propagation) neural network is proposed for parameter tuning of PID. BP neural network can achieve approximation of arbitrary nonlinear functions. Through neural network learning, the best combination of proportional, differential and integral coefficients is obtained to achieve the best control effect of BP_PID.