| To develop wind energy infrastructures in Taiwan as an alternative energy solution in place of partial fossil-fuel power generation will be helpful in improving national power supply stability as well as lowering the demand for imported fossil fuels. The stable operation of a wind farm relies on long-term collection of sensor data regarding marine weather and wind force conditions, including sea temperature, humidity, wind direction, wind speed, wave height, ocean current, and so on. With the aid of long-term collection of sensor data, the prediction model of local wind force condition is able to be constructed further to advance the efficiency of wind power generation. This paper concentrated on the issue of developing prediction model on top of GRU-based recurrent neural networks to predict local wind conditions such as wind speed and wind direction by means of long-term sensor data from the measurement tower which is 6 kilometers off Changhua’s coast and was built by Taiwan Power Company. |