Dynamic reverse droop power sharing in microgrid based on neural networks

Abstract

This paper presents a dynamic reverse droop power-sharing scheme to control frequency of micro grids based on neural networks. Micro grid is included undispatchable distributed generation (DG) such as wind turbine and dispatchable voltage source converter DG which is considered as auxiliary generation. Power sharing control strategies of distributed generation (DG) units without communication are based on the droop concept; while for the DG units in current control mode that applied in most renewable resources such as the wind, the droop control may not be implemented directly since the output of droop control is voltage amplitude and frequency. Also, the droop coefficients are constant and determined as a function of DG unit capacity. However, the generation capacity of the wind turbine is not constant. Therefore, in the proposed scheme first, the reverse droop method is provided for the wind turbine power sharing as its outputs are active and reactive power reference. Then, the dynamic coefficients are modified as a function of generalized regression neural networks. Finally, the control strategy is presented for coordinate control of wind turbine and auxiliary generator (AG) to ensure frequency stability of micro grid. The proposed control strategy is validated through extensive simulation results using MATLAB/Simulation software.