Resonance Mechanism And Control Strategies Of Power System Low Frequency Oscillation

Low frequency oscillation is one of the key problems which influence the security and stability of the interconnected system operation. In this thesis, the mechanisms which might induce the low frequency oscillation are thorough studied and analyzed, based on that, a set of effective analysis methods and control strategies for it are presented.A new method is presented in this thesis to improve the Prony algorithm. The idea of the method is to determine the order of Prony algorithm according to the principle of the minimum of the mean square deviation of fitting result, and fix the length of date window by the strategy of the combination of section analysis and synthesis analysis. By the method, the problem that analysis error by conventional Prony algorithm is large could be get rid of, when signal contains noise.Focusing on the problem that the performance of tradition power system stabilizer is influenced greatly by its parameters, a new algorithm to optimize the parameters of power system stabilizer is proposed in this thesis. The algorithm adopts objective function based on two eigenvalues, and the Particle Swarm optimized technology to be the calculation method to search for the optimal settings of PSS parameters. The PSS whose parameters is gained by the algorithm could effectively reduce the local and interarea modes of oscillations, and the robustness of system is obviously enhanced.An design method for the PSS based on the neural network sliding mode variable structure is proposed. The method adopts the algorithm of pole assignment to construct the sliding plane, at the meantime, adjusts the parameters of the controller on line by neural network to abate the dithering. The test results demonstrate that this PSS possesses excellent robustness and self-adaptive capability.In order to improve generalization capacity and convergence of feedforward neural network, a new training method for the neural network based on compound fitness particle swarm optimization is put forward. The results show that the new method can improve the performance of feedforward neural networks remarkably.This paper proposes a new design scheme to construct a SVC damping controller based on neural network, which consists of a neural identifier and a neural controller, to enhance its adaptability. A hyperbolic tangent type is used as the activation...