Application Of Improved LMS Algorithm In Identification Of Power System Low Frequency Oscillation Mode

With the development of industry and the improvement of people's living standard,The demand for electricity in the whole society is increasing.The interconnection of power grids and the long-distance with large-capacity transmission have become an inevitable trend of development.But it also creates the conditions for the occurrence of low-frequency oscillation of power system,and the application of more and more high-gain exciters increases the possibility of low-frequency oscillations.Low frequency oscillations can pose a serious threat to the stable operation of the power system.Therefore,judging whether a power system has low frequency oscillation is a hot issue in the field power system stability analysis.In this paper,adaptive filtering is applied to the identification of low frequency oscillations in power systems.Firstly,the definition,mechanism,analysis method and suppression measures of low frequency oscillation in power system are summarized.The structure and working principle of adaptive filtering are introduced.The Method of Steepest Descent,the least mean square algorithm and the recursive least squares algorithm are deduced and analyzed.The principle and flow chart of the adaptive filtering algorithm applied to the identification of low frequency oscillations in power system are also introduced.Build a New England 10-machine 39-node system.An adaptive filter based on least mean square algorithm is applied to the identification of low frequency oscillation.The effectiveness of the algorithm is verified.At the same time,it is pointed out that there is a contradiction between convergence speed and steady-state error in the fixed-step least mean square algorithm.The slow convergence speed of small step and the large steady-state error of large step lead to the decrease of recognition accuracy.This leads to a new variable step size LMS algorithm based on the improved Sinh function proposed in this paper.The algorithm also has a faster convergence speed while ensuring accuracy.When there is an outlier or dynamic data in the signal input to the adaptive filter,at this time,the adaptive filter based on the least mean square algorithm cannot accurately identify the information of the low frequency oscillation when encountering the disturbance.The algorithm has poor impact resistance.This paper introduces the idea of Maximum Correntropy Criterion into the algorithm.Making it resistant to shocks,It is possible to identify low frequency oscillations in the presence of disturbances.The reason why the Maximum Correntropy Criterion algorithm can resist shock is that a Gaussian kernel is used.The influence of the kernel width parameter of the Gaussian kernel on the algorithm is similar to the step factor.In order to ensure that the algorithm maintains the ability to resist shocks and speed up the convergence of the algorithm.In this paper,an improved kernel-wide Maximum Correntropy Criterion algorithm based on Tanh function is proposed.The algorithm has better robustness and faster convergence speed for the recognition of oscillation patterns.