The Direct Atomic Decomposition Method Based On Intelligent Optimization And Its Application In The Analysis Of Low-Frequency Oscillation

With the development of large-scale interconnection project,the low-frequency oscillation of power system has become a prominent problem affecting the security and stability of power grid,so the research of low-frequency oscillation pattern recognition algorithm has become one of the hot topics in power system.The ultimate aim of the study on low-frequency oscillation is to realize on line monitoring and control of it.The on-line monitoring system can realize the real-time monitoring of the oscillation in the power system.Once the low-frequency oscillation occurs,the system will quickly identify the mode information of the low-frequency oscillation.By analyzing the change of attenuation factor and other parameters,we can analyze the mechanism of oscillation,and then make the control strategy of low-frequency oscillation.Finally,the source or source of disturbance is removed by manual operation or computer control.In this way,we can find the low-frequency oscillation and control it in the bud,so as to minimize the harm caused by it,and ensure the safe operation of the power system.Online monitoring algorithm is an important part of the on-line monitoring system.In the past,scholars have studied the methods of low-frequency oscillation identification,which mainly focus on the accuracy and noise resistance of the method.However,the online monitoring system put forward new requirements for the detection methods,such as real-time,simple structure,stable algorithm,etc..In this paper,in order to explore the identification algorithm to meet the requirements of the online monitoring system of low-frequency oscillation,a new method is introduced to the detection of low-frequency oscillation,which is the atomic library sparse decompose(ALSD).However,when the greedy iterative algorithm(MP)is used to search for the best matching atom,the ALSD method uses too many nested iterations to make the recognition speed slower.And the ALSD method needs to decompose the signal into Gabor atoms which is easier to be dispersed,and then convert them into decaying sine atoms,the process will produce errors,so the accuracy of the method is low.In this paper,to solve the problem of speed and precision of this method,ALSD is combined with intelligent optimization algorithm,and then two improved methods are proposed.They are improved atomic library decomposition method based on particle swarm optimization(PSO-ALSD)and the improved atomic decomposition method based on chaos optimization(COM-ALSD).And the two methods are both use the direct damped sine atom to decompose the signal.In order to compare the performance of the method,this paper uses the ideal signal?simulink simulation signal of four machine two area system and PMU measured signal of a power grid in Central China to test these methods.The results show that the PSO-ALSD algorithm and COM-ALSD algorithm are better than ALSD algorithm and SVD-Prony algorithm in the identification accuracy,noise resistance,stability and resolution.In the real-time aspect,COM-ALSD algorithm performs best.Compared with other algorithms,its usage is reduced by an order of magnitude.Therefore,the COM-ALSD algorithm can meet the requirements of on-line monitoring system,and has some practical value.In addition,by using the COM-ALSD algorithm and the whole PMU measured data,this paper simulates the process of the low-frequency oscillation identification.And this paper explored how to determine the type of oscillation by the change of the oscillation frequency and damping coefficient,which can provide reference for the next step: the control of the low-frequency oscillation.