Methods based on ARIMA and Holt-Winters models for ultra-short-term electricity demand forecasting

Abstract

Abstract In this paper, several practical methods for modeling and forecasting load demand in the ultra-short term are presented. The proposed methods are: ARIMA and Holt-Winters, which are used for future day demand forecasting, some practical considerations for ultra-short term time windows are evaluated. Load demand data from the National Interconnected System of Ecuador with 30 minute granularity is used for the implementation. An evaluation of the two forecasting methods in two different time contexts is also performed to determine their effectiveness and accuracy in each scenario. The ARIMA method has proven to be highly effective, since it has obtained successful results in the two time contexts evaluated.