MODELLING WEEKDAY HALF HOURLY ELECTRICITY DEMAND

摘要

This paper reports a model to describe half hourly electricity demand during working days to use as input in simulation software and eventually for forecasting. Since the intra-day pattern differs according to the season, four different models were constructed. Similar procedures were conducted for each season. The daily cycles were described using Fourier series. In order to explain the association between demand and current temperature, linear and quadratic models were implemented. The relationships with past temperature and demand of the two previous days were analysed applying multiple regression models. Finally ARIMA models were used to explain the noise behaviour. The models describe satisfactorily the behaviour of the electricity demand. The average error of prediction one step ahead is lower than 1.6% for all the seasons.