SHORT-TERM ANALYSIS OF DISTRIBUTION SYSTEMS WITH WIND FARMS USING MULTI RESOLUTION ANALYSIS NEURAL NETWORKS

摘要

Since wind power generation as a major input of the power distribution system strictly relies on wind speed, stochastic wind fluctuations challenge an accurate prediction of the power generation. In this paper, an enhancement on Artificial Neural Network-based predicting model has been used to predict the major input of the system. The correlation between a smooth-dilated wavelet and wind speed samples has been calculated to decompose wind series into resolutions in which there is more regular patterns to be approximated. This predicting scheme was validated on the real wind (generation) data as well as consumption level of a real system to give a short-term prediction of the system inputs. The unbalanced three-phase load flow as a robust analysis tool determines system state variables under these predicted inputs. The results indicate the high accuracy of the method in predicting the system state variables from 3 hours to a day in advance.