When the improved droop control based on virtual impedance is adopted in island microgrid, the problem of inaccurate distribution of reactive power and reactive power circulation will still occur with the changing line impedance due to the fixed value of virtual impedance. To solve this problem, a virtual impedance prediction model based on partial least squares regression (PLSR) is proposed, which uses the line impedance value and the system impedance value before compensation to predict the virtual impedance value and realizes the adaptive virtual impedance, thus overcoming the problem in the improved droop control based on virtual impedance. There is no need to detect the real-time power value and circulation value, and the use of communication network is not required. Furthermore, from a comparison with the prediction results obtained by neural network models, it is proved that the prediction accuracy of the virtual impedance prediction model based on PLSR is better. At last, a simulation system of microgrid is constructed in MATLAB/Simulink to verify the adaptive virtual impedance, and simulation results show the superiority of the proposed model.

Microgrid  /  virtual impedance  /  multi-correlation  /  partial least squares regression (PLSR)