The Equivalent Modeling And Stability Analysis Of Microgrid Based On Neural Network

Micro-grid is a small power distribution system, which is composed of distributedgeneration, energy storage, energy conversion devices, power load monitoring andprotection devices. It can not only run well when connected to the external grid, but it canalso be operated isolatedly when the main grid doesn’t work. With the growing importanceof the micro-grid, getting the accurate model of the micro-grid arouses the specialattention of the researchers gradually. However, because of growing size of the networkwith the increase in the number of micro-grid and also the difficulty of the calculation ofthe interactive micro-grid system, getting an accurate model of the system is becomingmore and more difficult. On the other hand, first, the method of approximating thedynamics of the distribution networks using passive lumped loads will lack the sufficientaccuracy greatly, and second, the micro-grid requires accurate and fast analysis, so, basedon the previous researches on equivalent modeling method, a new equivalent modelingmethod using neural network is provided and also used in stability analysis in this paper.Firstly, to solve the great challenge of mathematical modeling in micro-grid systemanalysis, a new equivalent modeling method which uses neural network to equal localmicro-grid system structure is provided on the basis of researches on the self-adaptive ofneural network and traditional equivalent modeling method. And the related theoreticalanalysis is also carried out.Secondly, a micro-grid with photovoltaic cells and turbine is builded on Matlab andlocal network parameters are extracted. After neural network training, an equivalentmodule generated from it is used to take the place of the local network. We use thecomparison of the relevant data and the original network to verify the equivalent model.Finally, the stability analysis methods are introduced, and stability of the modelbuilded in the paper is analysised using this method. We also extracted the voltage andcurrent value of the original network and the equivalent network after adding thedisturbance and draw out the static and dynamic error curve before and after theequivalence. At last, a number of measurements are carried out with different fault types to verify the stability of the equivalent model in different disturbed states and also to providea new way for further analysis of the micro-grid model.