Effect of Volume of Historical Data on Short-Term Wind Forecasting Accuracy using Time-Series Analysis

Shivappa V. Sobarad and Suresh H. Jangamshetti


Wind speed forecasting, ARIMA, transfer function model, volume of historical data


This paper presents the outcome of a study conducted for measuring the forecasting accuracy with varying volume of wind speed data and increase in variables in the forecasting models. It is found that the forecasting accuracy depends on the volume of historical data and number of variables used to develop the model. Also the forecasting accuracy affected by forecasting horizon too. Wind speed and temperature data are found to be most effective in improving forecasting accuracy. The Transfer Function ARIMA model is developed using the historical data from July-2007 to July-2012. Initially only one year data is used to test the model and later model is tested using two, three, four and five years data. It is found that, as the volume of data increases, the forecasting accuracy also increases accordingly. In addition to this temperature data is also included in developing the forecasting model. It is observed that accuracy increases remarkably. The wind site selected for study is located at Basaveshwara Engineering College, Bagalkot, Karnataka, India. Wind and temperature data are measured using 50m-wind mast with interval of 10min between each data. The forecasting also done for the site located at Agricultural University at Bijapur, Karnataka, India for verification of the results. The proposed models can be used by utilities for planning, maintenance scheduling and load management of wind farms

Important Links:

Go Back