Research On Model Reduction Technology Of Grid-connected Inverter Network Based On Three-phase Droop Control

For a long time,the coherency equivalence method has been widely used to establish dynamic reduction models of large-scale synchronous generator networks.The advantage of this method is that the obtained reduced-order model maintains the nonlinear characteristics of the detailed model,which enables efficient and accurate analysis of large-scale network dynamics when large-scale disturbances such as short-circuit faults occur.Among them,the representative slow coherency algorithm has the advantages of being insensitive to the location and severity of the fault.However,due to model limitations and poor accuracy,traditional slow coherency algorithm is difficult to apply in the distributed generation network.To solve the above issues,this article takes the three-phase droop control grid-connected inverter as the basic model,based on the improved slow coherency algorithm,a coherency equivalence model with higher accuracy is established,and under different failure scenarios,the effectiveness of the order reduction method is verified.On the basis of summarizing the research status of model order reduction technology at home and abroad,this paper is organized into the following aspects.（1）The construction method of the three-phase droop control grid-connected inverter network mathematical model is studied,after obtaining the precise mathematical expression of each control structure and network structure in the inverter,combine the parts to get the complete model.The established mathematical model of the inverter is the basis of the full text research.（2）The implementation method of singular perturbation theory is studied,based on the multi-timescale characteristics of the three-phase droop control grid-connected inverter,dynamic processes on fast time scales are ignored,and the singular perturbation model is established under the slow time scale.The inverter is seen as a virtual synchronous generator through the model,and provides an effective way to establish a slow coherency model of the droop control inverter system,the application range of the slow coherency algorithm is expanded to the distributed power generation network based on the inverter.（3）An improved slow coherency algorithm is proposed.Based on the traditional slow coherency algorithm,the damping torque of the droop control inverter is considered,and the accuracy of the slow coherency algorithm applied to distributed power generation is increased through additional clustering.The accuracy of the slow coherency algorithm is further improved by combining the FCM clustering algorithm with the traditional slow coherency algorithm.It solves the problem of incorrect clustering results of traditional slow coherency algorithm when the desired number of clusters is greater than a certain number of clusters.According to the characteristics of the power supply in the slow coherency clustering area,a corresponding aggregation method is proposed to establish a reduced-order model combined with the improved slow coherency algorithm.This solves the problem that when the system is too complex,the simulation takes a long time and some research cannot be carried out.（4）Combining the research results of mathematical modeling and model reduction technology in this paper,in different scenarios,the simulation verification is carried out based on the network of 10 parallel droop control inverters.The reduced-order model without considering the coherency identification method,the improved coherency equivalent model,the traditional coherency equivalent model,and the full-order model are compared and analyzed under simulation,and simulation waveforms verify the effectiveness of this reduced order method and the feasibility of extending the application in different power supply models and systems of different scales.