NARX models for predicting power consumption of a horizontal axis wind turbine

Wind prediction is a key technique for seamless integration of large penetration of wind power into the power system. In this study, a nonlinear autoregressive model with exogenous inputs (NARX) is developed to predict the power consumption of a single wind turbine. The training data for the NARX models are collected from a 1.5MW wind turbine of a wind farm located at Northeast of China. The accuracy of models using different independent inputs including wind direction, speed as well as engine room position, is compared. The results show that the resultant NARX models are capable of capturing the dynamic characteristics of wind turbine power consumption, and the modelling accuracy with three inputs is better than of the model using only two inputs.