Original scientific paper
https://doi.org/10.37798/2017661-4107
Short-term power system hourly load forecasting using artificial neural networks
Ninoslav Holjevac
orcid.org/0000-0001-6570-2757
; Univesity of Zagreb Faculty of electrical engineering and computing, Zagreb, Croatia
Catarina Soares
; Univesity of Zagreb Faculty of electrical engineering and computing, Zagreb, Croatia
Igor Kuzle
orcid.org/0000-0001-8992-4098
; Univesity of Zagreb Faculty of electrical engineering and computing, Zagreb, Croatia
Abstract
Artificial neural networks (ANN) have been used for many application in various sectors. The learning property of an ANN algorithm in solving both linear and non-linear problems can be utilized and applied to different forecasting problems. In the power system operation load forecasting plays a key role in the process of operation and planning. This paper present the development of an ANN based short-term hourly load forecasting model applied to a real data from MIBEL – Iberian power market test case. The historical data for 2012 and 2013 ware used for a Multilayer Feed Forward ANN trained by Levenberg-Marquardt algorithm. The forecasted next day 24 hourly peak loads and hourly consumptions are generated based on the stationary output of the ANN with a performance measured by Mean Squared Error (MSE) and MAPE (Mean Absolute Percentage Error). The results have shown good alignment with the actual power system data and have shown proposed method is robust in forecasting future (short-term) hourly loads/consumptions for the daily operational planning.
Keywords
Artificial neural networks; Short-term load forecasting; Electric power system operation and planning
Hrčak ID:
199752
URI
Publication date:
15.12.2017.
Visits: 944 *