遗传优化的神经网络非线性自适应逆控制

Non-linear adaptive inverse control to neural network based on the genetic algorithm

针对BP网络收敛速度慢、易陷入局部极值等缺点,提出一种遗传算法优化神经网络权值的算法。利用该网络精确性和稳定性的优势,作为初始控制器,在线辨识对象的逆模型,然后与对象串联,实现了一种非线性系统的直接自适应逆控制策略。在对象特性未知的情况下,用BP网络作为辨识器对被控对象建模,并由辩识结果对控制器的参数进行在线调整。仿真结果表明该方法能够对非线性系统实施有效的控制。

Aiming to the BP network＇s flaws of low convergence speed and being ape to involve the local limit value, a weight algo- rithm of neural network optimized by genetic algorithm is given. Taking advantage of its accuracy and stability, the new algorithm identifies the inverse model of object online as the initial controller, and then cascades with object, fulfills a direct adaptive inverse control strategy of non- linear system. On the condition of unknown features of object, the new algorithm uses BP network as the identifier to model the controlled object, adjusts the parameter of controller online with the identifying results. The outcome of simulation proves that the new algorithm can bring the effectual control to non-linear system.