Wind speed estimation in Saudi Arabia using the particle swarm optimization (PSO)

Particle Swarm Optimization (PSO) is used in this paper to train a neural network to estimate wind speed 24 hours ahead based on the previous wind speed at 72 hours. Four years ofhourly wind speed data at Rowdat Bin Habbas, Saudi Arabia, between 2006 until 2009 are divided into three groups 50% is used for training, 25% for validation and 25% for testing. The validation data set is used to select the network architecture and other PSO user defined parameters. The testing data is used only to assess the generalization capability ofthe network on future unseen data that has never been used for training or model selection. Twenty four networks, each for wind speed at one future hour are used. Close agreements were found between the PSO predicted and measured hourly mean wind speed. For testing data set, the RSME varied from 0.0140 to 0.1826 at hours 14 and 18 while MBE from 0.1657 to 0.5550 corresponding to hours 1 and 8. Performance indicates thatthe proposed algorithm is viable for predicting wind speed.