On the Wind Speed and Power Prediction via New Hybrid Prediction Structure

Wind speed and power forecasting are the remarkable measures for wind farm management and secure integration into the electric power grid. Meanwhile, the accurate forecast of wind speed is important for the safety of renewable energy utilization, an exact wind power forecasting can highly assist the distribution and transmission system managers in preparing power system management. Based on the theories of Wavelet, some approaches such as feature selection, classical time series analysis, imperialistic competitive algorithm, enhance particle swarm optimization and artificial neural networks, the hybrid forecasting frameworks, the data modeling with imperialistic competitive algorithm (ICA)-multilayer perceptron (MLP), the data modeling with enhanced particle swarm optimization (EPSO)-MLP, the data modeling with GMDH and MLP-GMDH, are proposed to predict wind speeds and powers. Analogies of forecasting performance using various algorithm compounds are provided to investigate the contribution of diverse components in those hybrid frameworks. The results based on experimental cases of two depict that; the MLP-GMDH proposed hybrid forecasting frameworks is appropriate for the different precision requirements in wind speed and power predictions rather than the other methods. Contrary to the data modeling, two neural network frameworks, the combination of GMDH with MLP can improving the MLP with metaheuristic algorithms which are not statistically significant.

Keywords:

Wind speed and power forecasting , Enhance particle swarm , Imperialistic competitive algorithm , artificial neural network , Wavelet packet decomposition