A Wind Direction Forecasting Method Based on Z_Score Normalization and Long Short_ Term Memory

Abstract

The angle between wind direction and wind turbine affects the utilization efficiency of wind power generation. Ideally, the wind power generation efficiency is the highest when wind direction is perpendicular to the wind turbine. However, the wind direction is changing all the time, so it does not always keep perpendicular to the wind turbine. Therefore the efficiency of wind power generation can be improved by adjust the wind turbine perpendicular to wind direction. This paper proposes a wind direction forecasting method based on z _score normalization and LSTM. Z_ score normalization is used to preprocess data of wind direction, then the normalized data is feed to the LSTM neural network to train. The future wind direction is predicted by the trained LSTM neural network to adjust the angle of the wind turbine to make it as close as possible to be orthogonal with the wind direction so that maximize the efficiency of the wind. The experimental results show that the LSTM neural can predict the short-term wind direction angle exactly compared with benchmarks.