神经元PID控制器及改进算法的仿真研究

Simulation of PID Controller Based on Artificial Neural Cell and Its Improved Algorithms

针对常规PID控制器有着对过程的数学模型过于依赖的局限性,导致许多过程控制效果不理想的问题,根据人工神经元的自学习功能构造了基于神经元的PID控制器,对其学习算法加以改进。选取二阶惯性环节加纯滞后为控制对象,建立了数学模型,并进行计算机仿真及对这几种控制方法的控制效果加以比较。仿真结果表明,该控制器将神经网络和PID控制规律融为一体,既具有常规肿控制器结构简单、参数物理意义明确的优点,又具有神经网络自学习、自适应的能力,取得比常规PID控制器更好的控制品质。

Conventional PID controller depends on mathematics model to get good control performance. To pursue better effect, a type of PID controller based on artificial neural is proposed, and its improved algorithms are present. A second-order plant with deadtime which is common in industry nowadays is chosen as control object, mathematical model is created and realized through computer simulation. The result of simulation indicates that the controller is a combination of neural network and the PID rules. It has not only the simple structure and definite physical meaning as with a classical PID controller, but also the self-learning ability as neural net. What's more, this controller produces more preferable control quality than a classical PID controller.