Forecasting Hourly Electricity Demand Using a Hybrid Method


International Conference on Consumer Electronics and Devices (ICCED), London, Canada, 14 - 17 July 2017, pp.8-12 identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Volume:
  • Doi Number: 10.1109/icced.2017.8019981
  • City: London
  • Country: Canada
  • Page Numbers: pp.8-12
  • Yozgat Bozok University Affiliated: Yes


In the electricity sector, new sides have emerged with the development of technology and the increasing the electric energy need. Today, electricity has become a product that is bought and sold in the market environment. Forecasting which is the first step of plans and planning have become much more important and have been made mandatory for the market participants by energy market regulators. In this study, a short-term electricity load forecast is done for 24 hours of next day. Artificial Neural Network (ANN) and Particle Swarm Optimization (PSO) techniques are used for the forecast method in a hybrid form. The weights of ANN is updated by PSO in learning phase. Historical load consumption data, historical daily mean air temperature data and season are selected as inputs. Load data of 4 years on hourly basis are taken into account. Train and test data are considered as 3 years and 1 year, respectively. The MAPE error is found as 2.15 for one year period on an hourly basis.