Wind Speed Prediction for a Target Station using Neural Networks and Particle Swarm Optimization

In this study, artificial neural networks (ANN) and particle swarm optimization (PSO) were applied to predict the average daily wind speed measured at a meteorological tower installed June 2005 at the Greater Moncton Sewerage Commission, in Moncton, New Brunswick, Canada. Wind speeds were collected covering the period between June 2005 and December 2008. Five reference airport meteorological stations were used as input for the neural network and PSO. The artificial neural network modeling was done using the Matlab r neural network toolbox, while the PSO algorithm is an in-house program written in Matlab r. The daily wind speeds generated by the ANN model and PSO were compared with the actual measured data. It was found that with six months of input data, both the ANN and the PSO were able to predict the short term daily wind speed for the following 36 months at the target station. The PSO obtained a smaller error compared to the neural network. The PSO algorithm was also able to find the best combination of input variables automatically, while the ANN used manually-selected input variables.