An Empirical Study of Knowledge Discovery on Daily Electrical Peak Load Using Decision Tree

Abstract

Load forecasting plays an important role in effective energy management and system reliability. However, difficulty in load forecasting arises due to its nonlinear and irregular variation. This paper presents a case study in which the effect of weather elements on public utility system peak demand is first studied to find the most decisive weather index. The average temperature and previous load demand data are then selected as input variables to predict next-day peak load. The empirical results indicate that the forecasting performance can be improved by applying support vector regression machines.