Short Term Electricity Load Forecasting with Nonlinear Autoregressive Neural Network with Exogenous Variables (NarxNet)

Yazıcı İ., Temizer L., Beyca Ö. F.

in: Industrial Engineering in the Big Data Era, Fethi Calisir,Emre Cevikcan,Hatice C. Aladag, Editor, Springer, London/Berlin , Nevşehir, pp.3-513, 2019

  • Publication Type: Book Chapter / Chapter Research Book
  • Publication Date: 2019
  • Publisher: Springer, London/Berlin 
  • City: Nevşehir
  • Page Numbers: pp.3-513
  • Editors: Fethi Calisir,Emre Cevikcan,Hatice C. Aladag, Editor
  • Istanbul Technical University Affiliated: Yes


Electricity load forecasting and planning have vital importance for sup- pliers as well as other stakeholders in the industry. Forecasting and planning are relevant issues that they provide feedback to each other to increase the efficiency of management. Accurate predictions lead to more efficient planning. Many methods are used for electricity load forecasting depending on the characteristics of the system such as stationariness, non-linearity, and heteroscedasticity of data. On the other hand, in electricity load forecasting, forecasting horizons are important issues for modeling time series. In general, forecasting horizons are classified into 3 categories; long-term, mid-term and short-term load forecasting. In this paper, we dealt with short-term electricity load forecasting for Istanbul, Turkey. We utilized one of the efficient nonlinear dynamic system identification tools to make one-step-ahead prediction of hourly electricity loads in Istanbul. In the final, the obtained results were discussed.