Combined EMD-ELM and OS-ELM techniques based on feed-forward networks for PV power forecasting

Abstract

The grid integration of distributed photovoltaic (PV) generation needs to improve prediction technique so as to predict PV generation power. Two inventive and effectual method namely online sequential extreme learning machine (OS-ELM) and EMD-ELM forecasting technique are represented in this paper for short, medium and long term prediction of PV generation. Firstly the prediction performance of the combined empirical mode decomposition (EMD) and extreme learning machine (ELM) forecasting technique is compared with proposed forecasting technique. These algorithms are implemented with the single hidden layer feed forward neural network (SLFN) for a real time PV model in MATLAB software. The forecasted results of each section are superimposed and compared with these two forecasting techniques to evaluate prediction accuracy of the proposed forecasting technique. The simulation result shows that the OS-ELM forecasting technique gives better generalization performance and higher prediction accuracy than EMD with ELM forecasting technique. These models can help to regulate the generation of grid energy management, schedule the power generation as well as support the integrated power control, which is necessary for the safety and maximum operation of power system.