基于神经网络的同步发电机励磁反步控制

针对同步发电机励磁系统严重的非线性以及参数非精确可知的特点，提出了一种基于神经网络的反步(Backstepping)控制方法，实现了多目标量的励磁控制。神经网络的运用解决了反步控制与励磁系统不满足匹配条件的矛盾。系统设计采用了特殊的在线权值修正算法，不需要离线学习阶段。由稳定性证明可知：控制算法中各参数的设计并不要求被控对象参数精确可知，系统具有较好的鲁棒性。仿真结果表明：该控制方法与PID控制相比具有超调量小、调节速度快等优点。

BACKSTEPPING CONTROL OF THE SYNCHRONOUS GENERATOR BASED ON THE NEURAL NETWORKS

Aimed at the characters of serious nonlinear and unknown or uncertain parameters, a Backstepping control method based on neural networks is proposed to realize multi-object generator control. Neural networks are used to solve the contradiction of Backstepping control and unmatching of systems. A special weight online tuning method is been proposed in the system, and an off-line training phase was not required. According to the proof of stability, the method did not require the system parameters to be exactly known, and the system was robustness. The result of simulation verified effetely of the proposed method that as small overshoot, short tuning time and etc.