基于神经元的自适应预测PID控制

提出了一种采用神经元的自适应预测PID控制方案。神经元具有很强的自学习和自适应能力 ,它可根据系统的误差并通过一定的学习算法来不断修正控制器的参数 ,使控制器能够适应受控对象结构参数以及环境的变化。因此采用单神经元构成自适应PID控制器 ,将神经元与PID控制结合 ,对PID参数进行在线寻优、自校正 ,使PID控制能有效地对付一些较复杂非线性被控对象 ,特别是难于用传统方法建模的被控对象。同时为了提高系统的快速跟踪和抗干扰能力 ,采用了动态自适应神经元 (APE)对非线性系统进行预测 ,即用神经元建立起非线性系统的预测模型 ,预测系统的未来输出 ,从而提高了控制系统的控制品质。同时详细介绍了该控制系统的自适应控制算法。仿真结果表明 ,这种自适应控制方案切实可行 ,其控制品质明显优于常规PID控制 ,且具有较强的鲁棒性 ,达到了良好的控制效果。

An Adaptive Neuron PID Control Based on the Optimal Prediction

An adaptive neuron PID control system based on the optimal prediction is presented. Neuron has strong ability of self-learning and self-adapting. It can correct controller parameters continuously according to the system errors by certain algorithm so that the controller can adapt itself to the parameter changes of object structure or environment. An adaptive neuron was used as a controller with PID control structure. The parameters of the PID controller can be adjusted on-line, or tuned by itself. Thus this controller can proceed some complicated non-linear controlled objects effectively, especially, the objects hardly modeled by traditional method. In order to enhance the system abilities of tracing quickly and disturbance resistance,an dynamically adaptive neuron is employed to predict the process on-line. That is, by using neuron to get the non-linear prediction model and predicting system's output, the controlling quality of system will be improved.Meanwhile, the on-line tuning methods for the neuron networks in the control system are described. The result of simulation shows that the presented adaptive neuron PID control system is practical and has high robustness.