An Investigation into the Applications of Machine Learning Algorithms on Wind Speed Prediction

Abstract

In recent times, wind energy is a highly demanded source of renewable energy today. Consequently, global demand for wind energy has increased and thus the construction of wind turbines. However, wind turbines are often met with unfavorable conditions such as highly erratic and variable wind speeds or even storms. This has further consequences such as greatly reducing the efficiency of wind turbines and leaving its body damaged which is economically unfavorable. Particularly, wind speed prediction is a steady variable to consider while looking for viable options to increase the power generation from wind turbines. This paper aims to assess the performance of various machine learning models in time-series wind speed prediction. My hypothesis is that among the machine learning models tested, Random Forest Regression will outperform the others in predicting wind speed. After training and testing the data, I found out that Random Forest Regression had the best performance with a mean squared error of 5.64 and mean absolute error of 1.81. It also had the highest coefficient of determination of 0.68 and supported my hypothesis. Thus, these results show how machine learning models are reasonable tools for wind speed prediction as well as that Random Forest Regression can be used for real-time wind speed prediction after some hyper parameter tuning. This has major implementations as the model can be used to increase the efficiency of wind turbines, improve their safety and help in maintenance planning.