Performance Analysis of Short and Mid-Term Wind Power Prediction using ARIMA and Hybrid Models
Ashoke Kumar Biswas, Sina Ibne Ahmed, Temitope Bankefa, Prakash Ranganathan, Hossein Salehfar
Abstract
Due to the high market penetration of wind power, efficient prediction methodologies are of paramount importance to promote wind power generation in the electricity market against more secure and dispatchable energy sources. This paper has proposed mix of regression and machine learning methods, such as Auto-Regressive Integrated Moving Average (ARIMA), Random Forest (RF), Bagging Classification and Regression Trees (BCART), and two hybrid models of ARIMA-RF and ARIMA-BCART to forecast one, two, and seven days of wind power generation. The prediction relies on weather data such as wind speed, wind direction, air temperature, air pressure, and density at hub height. The preliminary results indicate that ARIMA-RF and ARIMA-BCART aids in improving forecasting accuracy (i.e., NMAE 18%-26%) over standalone forecast mode of ARIMA.