[Dissertation] Power Output Prediction For Wind Farms With Machine Learning
Masters dissertation by Nkosinathi, African Institute for Mathematical Sciences
Members
Nkosinathi Hlophe, AIMS
Supervisor(s)
- Dr. Vukosi Marivate, University of Pretoria
- Dr. Bubacarr Bah, AIMS
Abstract
Wind power generated by wind farms depends on meteorological conditions such as wind speed, temperature, wind direction and many more. Wind power generated is proportional to the cube of the wind speed up to a certain level that is why wind speed prediction is important. Support vector regression and Long Short term Memory (LSTM) are two models that were developed to accurately predict wind speed at 3, 6, 12 and 24 hours ahead. The data used was transformed using z-score standardization and all the computation are based on the transformed data. In this work we show the performance of the two models comparing them to a persistence model which is used as a benchmarking model. The LSTM on overage performs better than the other two models with a root mean square error of about 0.6 for 24 hour forecasting.