Research On Characteristics And Controlled Islanding Strategy Of The Out-of-step Power System Based On Multi-source Information

As the inevitable trend of the power system development,the interconnection of regional power grids has many operational benefits.But with the deepening of China’s power grid interconnection degree and increased sophistication of "source-grid-load" power environment,the traditional security stability control now faces new challenges.Large disturbances lead to local failure.If not taking timely measures or handling improper,it may lead to major blackouts of the entire system.As the last defence of "three defensive lines",out-of-step separation can effectively prevent the collapse of power system.The development of the wide-area measurement system and communication technologies has led to a new chapter in the study of out-of-step separation.How to use the multi-source information including wide-area information to identify the out-of-step state of the power grid,and to determine the optimal splitting surface are the key research problems of controlled islanding control.In this paper,a comprehensive strategy is put forward for controlled islanding technology such as out-of-step identification and splitting surface searching.The main work of this paper is summarized as follows:In identification of out-of-step state resesrch,a method of out-of-step identification based on prediction of power angle,rotation speed difference and unbalanced power trajectoriese is proposed.After the fault clearance,a short-term prediction of disturbed trajectories is implemented through self-memory prediction model and trigonometric function curve fitting model.Then,the phase plane is formed by the prediction of the disturbed trajectories and the dimension reduction.According to the geometric characteristics of the trajectory in the phase plane,the real-time power grid out-of-step state is discriminated.Response data before the power system is out of step is used.It is not necessary to know the system model,so this method is suitable for different operating modes.Compared with traditional methods such as maxdifference of generators’ power angle,the out-of-step identification time of this method is much earlier.In splitting surface selection resesrch,a method of searching the optimal splitting surface based on graph theory and particle swarm optimization is proposed.Based on coherent generator groups,the power grid structure is updated by graph theory method,which can ensure coherent constraints and reduce the search space.On the updated network,it is equivalent to search the optimal partition instead of searching the optimal splitting surface by particle swarm optimization for the purpose of minimizing the net power imbalance within islands.This method improves the fitness function and conbines community structure theory in consideration of partition rationality.Strong adaptability to different instability modes and reduction in the possibility of losing the optimal solution are virtues of this method.It is worth mentioning that this method provides an idea for the application of intelligent optimization algorithm in solving the problem of splitting surface search.In controlled islanding stragety research:First,a new out-of-step splitting criterion based on measured and predicted trajectories is presented for the purpose of avoiding the adverse effect of prediction error and improving the recognition speed based on the measured trajectory.Then,a new method based on machine learning is proposed in order to improve the utilization ratio of the multi-source information and the recognition speed of transient instability.Finally,an online method to develop controlled islanding strategy based on multi-source information is put forward.Through the simulation in New England 39 node system,the efficiency of the proposed controlled islanding strategy is verified.