EFFECT OF VOLUME OF HISTORICAL DATA ON SHORT-TERM WIND FORECASTING ACCURACY USING TIME-SERIES ANALYSIS

摘要

This paper presents the outcome of a study conducted forrnmeasuring the forecasting accuracy with varying volumernof wind speed data and increase in variables in thernforecasting models. It is found that the forecastingrnaccuracy depends on the volume of historical data andrnnumber of variables used to develop the model. Also thernforecasting accuracy affected by forecasting horizon too.rnWind speed and temperature data are found to be mostrneffective in improving forecasting accuracy. The TransferrnFunction ARIMA model is developed using the historicalrndata from July-2007 to July-2012. Initially only one yearrndata is used to test the model and later model is testedrnusing two, three, four and five years data. It is found that,rnas the volume of data increases, the forecasting accuracyrnalso increases accordingly. In addition to this temperaturerndata is also included in developing the forecasting model.rnIt is observed that accuracy increases remarkably. Thernwind site selected for study is located at BasaveshwararnEngineering College, Bagalkot, Karnataka, India. Windrnand temperature data are measured using 50m-wind mastrnwith interval of 10min between each data. The forecastingrnalso done for the site located at Agricultural University atrnBijapur, Karnataka, India for verification of the results.rnThe proposed models can be used by utilities forrnplanning, maintenance scheduling and load managementrnof wind farms.