基于在线学习RBF神经网络的汽门开度自适应补偿控制方法

Online learning RBF neural network-based adaptive compensative control scheme for governor system

汽门控制对于提高电力系统暂态稳定具有重要作用。为了提高汽门系统的控制性能,提出了基于在线学习RBF神经网络的汽门开度自适应补偿控制方法。首先,根据逆系统方法分析了被控汽门系统的可逆性、推导了被控汽门系统输出的α阶导数和伪控制量之间的误差,并设计了用于补偿此误差的在线学习RBF神经网络。然后,基于Lyapunov稳定性理论设计了RBF神经网络的在线学习算法,证明了闭环系统跟踪误差和RBF神经网络权值估计误差的一致最终有界性。所提出的控制方法仅需被控汽门系统很少的先验知识,而无需其精确数学模型,并且用于自适应补偿控制的RBF神经网络无需离线训练过程。最后,针对典型的单机无穷大汽门控制系统进行了数值仿真。仿真结果表明,所提出的控制方法较传统的非线性最优控制方法能明显提升电力系统的暂态控制性能。

Control of the governor system has a great impact on the transient stability of power systems. Aiming at improving the control performance of the governor system, an online learning RBF neural network- based adaptive compensative control scheme is proposed in the paper. Firstly, the reversibility of the controlled governor system is analyzed based on the inverse system method, the error between the α-th derivative of the system output and the pseudo-control is deduced, and an online learning RBF neural network is designed to compensate for the error. Then, an online learning algorithm for the RBF neural network is designed based on the Lyapunov principles, and the uniform ultimate boundness of the tracking error of the closed loop system and the weight estimation errors of the RBF neural network are strictly proved. The proposed control scheme does not require the accurately mathematical model of the controlled governor system but only a litter prior knowledge. Moreover, the RBF neural network for adaptive compensative control does not need offline learning phase. Finally, some simulations are conducted for a typical single-machine-infinite-bus governor system and simulation results show that the proposed control scheme can greatly improve the transient control performance of the power system than the conventional nonlinear optimal control scheme.