Abstract：In order to improve the accuracy and stability of short-term wind direction prediction，a wind

direction prediction method based on lidar wind data and an improved nonlinear echo state network（NESN）model was proposed.First of all，wind direction data 100 meters ahead of the wind turbine was obtained by laser wind detection radar. Secondly，the multivariate polynomial function was used to construct the nonlinear relation of the internal state of the reserve pool，the order of the weight matrix and the complexity of model calculation were

reduced. Finally，the prediction model was established and the simulation prediction was carried out on different

lidar data sets. The results show that compared with the nonlinear echo state network and adaptive neuro fuzzy

inference system（ANFIS），the mean absolute error（MAE），root mean square error（RMSE），normalized mean

absolute error（NMAE）and normalized root mean square error（NRMSE）of the improved NESN model are

significantly reduced，and the prediction accuracy and stability are improved. The accuracy of the wind turbine

alignment the wind direction is improved and the mechanical loss of yaw is reduced.

      Key words：laser wind measurement radar；wind prediction；nonlinear echo state network（NESN）；yaw