基于BP神经网络的自适应自抗扰控制及仿真  

Adaptive Active Disturbance Rejection Control and Its Simulation Based on BP Neural Network

作　　者：齐晓慧[1] 李杰[1] 韩帅涛[1] 

机构地区：[1]军械工程学院无人机工程系,河北石家庄050003

出　　处：《兵工学报》2013年第6期776-782,共7页Acta Armamentarii

基　　金：装备预研基金项目(9140A25070509JB3405)

摘　　要：针对被控对象参数变化大而快、外扰严重且不确定的系统,参数固定的扩张状态观测器(ESO)存在"总和扰动"估计精度降低、控制效果较差的问题,提出了一种基于BP神经网络的自适应自抗扰控制器(ADRC)。分析了引入自适应ESO的意义,剖析了ESO的结构,利用BP神经网络在线调整ESO参数并将这个自适应ESO嵌入到ADRC.仿真结果表明,改进的ADRC较常规ADRC具有扰动估计精度更高、控制量振荡幅度更小以及鲁棒性、抗干扰性更强的优点。Because the fixed parameters of the extended state observer （ESO） reduce the estimation precision of ＂total disturbance＂ and control effect for the systems which the parameters of the controlled objects change largely and fast or there being serious and uncertain outside disturbance, an adaptive active disturbance rejection controller （ADRC significance of introducing adaptive ESO as well ESO which parameters are adjusted online by Simulations show that the quantity, high robustness imp and ） based on BP neural network was proposed. The as the structure of ESO was analyzed, then the adaptive means of BP neural network was applied to ADRC roved ADRC has higher estimation precision, smaller range of controlling anti-interference performance compared with conventional ADRC

关 键 词：自动控制技术 自适应扩张状态观测器 自适应自抗扰控制器 BP神经网络 

分 类 号：TP273.2[自动化与计算机技术—检测技术与自动化装置]