Analysis Of Low Frequency Oscillation In Power System Based On Prony Method

Low frequency oscillation is common problem that endangers the safe and stable operation of power system, and it belongs to the category of small disturbance stability problems. When it is serious, low frequency oscillation even leads to the collapse of the whole system. With the power system development’s speed being accelerated day by day, and network scale being continually expanded, low frequency oscillation problem becomes more and more prominent. How to effectively identify specific modal information of the low frequency oscillation signal of power system, in order to take further effective suppression measures, is a topic of great significance.Prony analysis is the general method to calculate signal information. It is based on a linear combination of the exponential attenuation to fit uniformly-spaced sampling data, and can effectively calculate amplitude, phase, frequency and attenuation of the signal of low frequency oscillation. It is an analysis method that uses linear equations to solve nonlinear problem. The solution of characteristic equation’s coefficient is the core of the Prony algorithm and its traditional calculative method is using normal equation form. The traditional calculative method’s order number of the equations is large, which leads the coefficient matrix to be singular, the numerical condition to be poor so that affect the solution of equation. Therefore, a method based on the residual error iteration is proposed. The practical verification shows when this method’s order number is just slightly larger than or even equal to the effective rank, the signal parameters will be obtained effectively. The precision of solution is high, and numerical stability of equation’s solution is good.The Prony method based on iterative residuals need to construct normal equation, and calculation is large. Moreover, when the order number is selected to be the effective rank, the fast attenuation component of the signal will be not obtained effectively. Therefore, the Givens iterative method is proposed to compute the signal. This method doesn’t need to calculate the normal equation and can compute the coefficients directly. When the order number is equal to the effective rank, whether the signal contains fast attenuation component or not, the Givens iterative method can calculate the parameters of signal effectively.The example analysis verifies that whatever the order number is larger than the effective rank or equal to it, the Givens iterative method can be efficacious in calculating signal parameters, regardless of the signal includes the fast attenuation component or not.