基于确定学习的电力系统稳定性控制

Stability Control for Power Systems Using Deterministic Learning

传统自适应控制方法存在计算过于复杂、控制器奇异性的问题。为了解决这些问题,采用了一种简单的自适应控制方法,以保证电力系统中所有信号的有界性和有限时间内跟踪误差的收敛性。主要思想是将控制过程中的稳定闭环系统分解为一系列具有适当状态变换的线性时变扰动子系统,利用递归设计方法,建立了径向基函数神经网络的部分持续激励条件,保证了扰动子系统的指数稳定性。因此,在闭环信号循环轨道的局部区域内实现了闭环系统动力学的精确逼近,在闭环反馈控制过程中所学习的知识可以被重复利用,从而避免了在自适应控制过程中大量重复训练的过程。仿真结果表明了该方法的有效性。

In order to overcome the problem of complicated control,the possible controller singularity of traditional adaptive control method,a simple adaptive control method is used to ensure that the boundedness of all the signals in the power system and the convergence of the tracking errors in finite time.The main idea of control proposes is that the stable closed-loop system is decomposed into a series of linear time-varying disturbance subsystem with the appropriate state transformation.Using the recursive design method,radial basis function(RBF) neural network of the partial persistent excitation condition is established,which guarantees the exponential stability of disturbance subsystem.Therefore,the exact approximation of the closed-loop system dynamics is achieved in the local region along recurrent orbits of the closed-loop signals,and the knowledge learned in the closed-loop feedback control process can be repeated.