Wind Speed Prediction Based on Support Vector Regression Method: a Case Study of Lome-Site

Abstract

This study aims at predicting the hourly mean wind speed using a Support Vector Machine (SVM) based on a regression (SVR) model. The SVM for regression is part of the machine learning supervision for prediction methods that have proven very effective in recent decades. A linear programming method in Python compiler was used to implement the SVR algorithm. The wind speed data were retrieved from the meteorological center in Lome. The simulation of the model has been performed and compared to the measured data. Significant results were obtained with a mean squared error (MSE) of 0.128, and the coefficient of determination (R2) of 0.962. The results have indicated that a real time prediction of the wind speed in prior could be achieved with a consistent modelling based on the hourly average of the wind speed when using the SVR model, which may ultimately contribute to an efficient wind energy generation.