Abstract
As the electricity market is growing, the need for
accurate Short Term Load Forecasting (STLF) is increasing.
Electrical grid operators require STLF to plan schedules for
power generation plants. With the introduction of intermittent
renewable resources, the stakes are now even higher. Developed
countries have been fortunate in this regard as most of the
research on STLF focused on these countries and developed
highly accurate models. There is now a need to focus on
developing countries as these are substantial energy markets
with thriving economies and high population growth rates.
With the 43rd largest economy by GDP and 6th largest nation
by population, Pakistan is one such country. As the energy
demand of Pakistan is increasing, there is a need to understand
the energy demand patterns of its citizens better. PRECON is
an electricity consumption dataset of residential buildings in
Pakistan that can help in this regard. In this paper, we present
preliminary results of applying STLF techniques on PRECON.
These initial results show that Multiple Linear Regression and
Support Vector Regression perform better than Artificial Neural
Network ambient temperature and autoregressive attributes as
input variables. The results also discuss various performance
metrics, such as ME, RMSE, and MPE. The results show a
unique phenomenon, load shedding, not experienced in developed
countries.