Transient Stability Control Of Power System Based On Distribution Chart Sign Learning

With the increase of renewable energy integration,the stable operation of the power system has been faced with new challenges.In the past,the more mature methods in the transient stability assessment of power system were energy function methods and time domained simulation methods,but the increasing integration of power generation forms such as wind power to the grid has made the power grid increasingly complex,and traditional methods alone cannot adapt to modern power grid structure.Since the topology of the power system cannot be considered in the traditional method,node distribution information and other specific data,the limitations of matching with the system are gradually exposed.Based on the improved graph neural network,this paper conducts an in-depth study on the online evaluation of transient stability of wind power system and the collaborative control strategy before and after instability,and the research content is as follows:(1)Aiming at the problem of transient stability control strategy of power system with wind power,through the research on the application of the previous single control strategy in the field of transient stability of power system,this paper proposes an effective transient stability collaborative control strategy,which skillfully transforms the optimization constraint of collaborative control into a two-stage and three-layer robust optimization mathematical problem,and its economy and effectiveness is verified through examples.(2)In view of the problem of online evaluation of transient stability of power systems with wind power,the algorithms used in previous research cannot consider specific data such as the topology of the power system and node distribution information.In this paper,a transient stability evaluation method based on distribution graph signature learning is proposed,which replaces the pooling module that cannot contain the distribution information of bus data in the traditional graph neural network.The work in this paper establishes a predictor for evaluating transient stability,performing input feature selection,graph structure construction and modeling,and compares multiple evaluation indicators with other machine learning methods,which verifies that the evaluation method has good generalization ability in the face of new topologies.(3)Aiming at the problem of transient stability collaborative control of power systems including wind power,the previous online evaluation of transient stability based on traditional machine learning did not solve the problem of how to use the prediction results for transient stability control.In this paper,a collaborative control transient stability model based on distribution graph signature learning is constructed,and a collaborative control strategy model based on the idea of "offline training,online evaluation and control" is proposed.Through the cases,it is verified that the proposed model can achieve accurate evaluation and effective control of transient stability after failure,improving the safety and stability margin of the system,and show that the model can still achieve effective control of transient stability under the condition of system topology changing.The work done in this paper can improve the ability of the online transient stability assessment model to obtain node distribution information data,and the developed transient stability prediction and evaluation collaborative control system is not only conducive to the connection of large-scale wind power grids,but also provides new ideas for reducing the cost of transient stability control of power system and improving the transient stability margin of power system.