Support vector regression based electricity peak load forecasting

Abstract

Electricity load demand forecasting is integral for planning and execution of various projects vital to urban development. It is highly essential for safe operation of power systems and to prevent power failures in future. This paper discusses the use of machine learning strategies, specifically, Support Vector Machines, for predicting future peak load demand based on historical data. Support Vector Regression is implemented here using RStudio software package. The dataset used in this study includes a daily record of peak load consumption and the corresponding temperature and relative humidity for three consecutive years (2014-2016). Evaluation results clearly show the effectiveness of support vector regression for peak load prediction.