Short-termwind-speed-to-wind-power Forecasting Using A Hybrid Of Particle Swarm Optimization And Artificial Neural Networks

Abstract

Renewable energy has gained momentum over the last decade, with wind power technology leading by the number of projects implemented worldwide and academia as research[1]. Worldwide the growth of wind energy generation[2] has increased tremendously over the last 10 years, for instance, in 2000, installed capacity was 60GW and in 2010 it was 160GW. Furthermore, by 2015 the total global installation to 433GW. With that kind of growth, energy sector professionals have had to keep pace finding better, reliable and efficient ways in the management of wind power farms and turbine designs that are cost effective. Numerical Weather Prediction software’s (NWP) systems and meteorological tools are too expensive to acquire and maintain, therefore, these constraints have necessitated the development of accurate prediction system that are simple, fast and cheaper that can be used by system planners, power regulatory experts and the academic community through research. Short term wind speed to wind-power predicting model is the main object for this research , the model will be optimized using a hybrid of particle swam optimization (PSO) and neural networks. Matlab modeling environment has been used extensively in this research work.

The hybrid prediction model performance will be evaluated, see chapter 4 and obtained the following values, using the mean square errors (MSE) of 0.26, the mean average error(MAE) of 0.62 and the mean average percentage error(MAPE) of 18.2.