Short Term Wind Speed Forecasting using Hybrid ELM Approach

Objectives: To enrich the accuracy of the wind speed forecasting to calculate wind power generation connecting to grid with the help of machine learning algorithms. Methods/Statistical Analysis: A novel hybrid method, Persistence - Extreme Learning Machine (P-ELM) algorithm is proposed. It uses the features of both Persistent and Extreme Learning Machine algorithms. The historical data of meteorological parameters viz. pressure, temperature and past wind speeds are considered as input parameters and wind speed for the next instant as the output, measured at one hour interval for a period of a month are used for the study. Findings: The forecasting is carried out for three areas Guntur, Vijayawada and Ongole of Andhra Pradesh state, India for winter, summer and rainy seasons. The performance of the P-ELM is evaluated in terms of Mean Absolute Error (MAE) and Root Mean Square Error (RMSE). From the results obtained it can be concluded that the P-ELM algorithm leaves behind the fundamental Persistent and Extreme Learning Machine algorithm not only in terms of error metrics but also in simulation time. The entire simulation is carried out with the help of MATLAB 2013a software. Application/Improvements: This technique is very much helpful in real time forecasting of wind speed to avoid the high level of uncertainty involved in wind generation, there by increases the power system security and stability.