One-day-ahead electricity demand forecasting in holidays using discrete-interval moving seasonalities. (15th September 2021)

• Record Type: Journal Article
• Authors: Trull, Oscar; García-Díaz, J. Carlos; Troncoso, Alicia
• Is Part Of: Energy. Volume 231(2021)
• Publication Date: 2021-09-15
• Subjects: Time series -- Forecasting -- Electricity demand -- Anomalous load; Power resources -- Periodicals; Power (Mechanics) -- Periodicals; Energy consumption -- Periodicals; 333.7905
• DOI: 10.1016/j.energy.2021.120966
• Languages: English
• ISSNs: 0360-5442

Abstract

Abstract: Transmission System Operators provide forecasts of electricity demand to the electricity system. The producers and sellers use this information to establish the next day production units planning and prices. The results obtained are very accurate. However, they have a great deal with special events forecasting. Special events produce anomalous load conditions, and the models used to provide predictions must react properly against these situations. In this article, a new forecasting method based on multiple seasonal Holt-Winters modelling including discrete-interval moving seasonalities is applied to the Spanish hourly electricity demand to predict holidays with a 24-h prediction horizon. It allows the model to integrate the anomalous load within the model. The main results show how the new proposal outperforms regular methods and reduces the forecasting error from 9.5% to under 5% during holidays. Highlights: A novel electric load forecasting model for anomalous load. Use of Holt-Winters models with discrete-interval moving seasonalities. Analysis of the behaviour of the electricity demand during holidays and bridges. Reported error results of 4.5% for the hourly electricity load in Spain. Comparison of prediction accuracy with other state-of-the-art forecasting methods.