工业过程控制算法设计与仿真

Design and Simulation of the Control Algorithm Based on Smith-PIDNN

传统的Smith预估器在理论上解决了纯滞后系统控制问题,但由于它必须依赖被控对象精确的数学模型,在实际应用中存在局限。要克服Smith算法的局限性,采用一种具有比例、积分和微分神经元的PID神经元网络(PIDNN)与Smith预估器结合的算法,利用Smith对纯滞后系统进行预估补偿以及PID神经元网络的自适应、自学习和在线自调整控制器的参数的功能,解决大纯滞后系统的控制问题。仿真结果表明该Smith-PIDNN算法简单,稳定且收敛速度快,能够很好地改善大纯滞后系统的性能指标。

The traditional Smith predictor can solve the problem of time delay control system in theory. It must re- ly on the accurate model of the controlled object, so there are some limitations in practice applications. This paper a- dopted a neural network containing proportion neural unit, differential neural unit and integral neural unit, and com- bined it with Smith predictor. By using Smith predictor as a compensator for the time delay system and the neural net- work functions of self - adaptive, self - learning and on - line self - tuning of control parameters, this algorithm, called Smith -PIDNN, is well available for the large time delay control system. The simulation results show that this Smith - PIDNN algorithm has the characters of simple, stable and fast convergence, and can largely improve the controlling effect and performance of the large delay system.