A Design of Short-Term Load Forecasting Structure Based on ARIMA Using Load Pattern Classification

Abstract

This paper suggests a short-term electricity load prediction structure for collection of most efficient energy operating plan by predicting the next day’s demand from separately measured past electricity demand. The present short term load prediction methods predict the next day’s electricity expenditure using a probability method, time-series method or a nerve network model without separation between similar data in a traditional way or uses a standard method of season separation or day separation. However if data separation is not done, fluctuating parameter that affects load like season and day cannot be responded to and in the case of the latter there is a problem caused due to the recent increased fluctuation of climate data - separation without similarities arises and so the prediction rate decreases. This paper separated collected electricity expenditure into months, holidays, Monday, Saturday and weekdays by high similarities and short term electricity load prediction is carried out based on time-series prediction method ARIMA. Also, to evaluate the accuracy of the prediction simulation will be done based on the collected data.