【摘要】 基于模糊神经网络的电力系统稳定器具有适应电力系统非线性,且不依赖电力系统数学模型的特点,针对模糊神经网络隶属度函数的中心参数选取问题,提出了一种基于极大熵原理优化模糊神经网络的设计方法。该方法利用一个最优化的目标函数导出中心向量和宽度的学习算法,改善了网络的回归能力和泛化能力。针对电力系统发生的低频振荡问题,提出了一种基于熵优化模糊神经网络电力系统稳定器的设计方案。该方案避免了控制器对系统精确数学模型的依赖,利用神经网络的学习能力,在线自动生成训练样本,实现了电力系统的实时控制。仿真结果表明,提出的电力系统稳定器控制方案可以显著地提高被控机组的稳定性及电力系统的动态性能。更多还原

【Abstract】 The power system stabilizer based on FNN(Fuzzy Neural Network) adapts well to the nonlinearity of power system and does not rely on the precise mathematical model. A FNN design method based on maximum entropy principle is proposed to optimize the center parameters of its membership functions,which uses an optimized objective function to deduce the learning algorithms of center vector and width,improving its regression and generalization ability. For the lower - frequency oscillation of power system,a power system stabilizer design based on entropy optimization FNN is proposed, which, independent of the precise system model,fully uses the learning ability of neural network to generate on - line training samples and implement the real-time control of power system. Simulation results show that the designed power system stabilizer enhances significantly the stability of controlled generator and the dynamic performance of power system.更多还原