Conference Paper

Daily load forecasting using recursive Artificial Neural Network vs. classic forecasting approaches

Electr. Power Eng. Dept., Politeh. Univ., Timisoara, Romania
DOI: 10.1109/SACI.2009.5136297 Conference: Applied Computational Intelligence and Informatics, 2009. SACI '09. 5th International Symposium on
Source: IEEE Xplore


The aspects presented in the paper refer to the recursive artificial neural network (ANN) architecture for short term daily load forecasting. The paper emphasizes the importance of choosing the right training set used to teach the recursive ANN (RANN). Using specific data from the Banat region (situated in South-Western Romania), some daily load forecasts based on the proposed method are presented, analyzed and compared to other forecast methods. The results show that the RANN method provides a better load forecast that the traditional methods. On this basis, many useful recommendations are outlined.

8 Reads
  • [Show abstract] [Hide abstract]
    ABSTRACT: Within the paper, the authors deal with electric substation ancillary services power consumption analysis. They are focusing on load curve numerical processing. A real case study is discussed, Timisoara 400/220/110 kV substation, the consumers being supplied from the autotransformer tertiary winding. The parameters specific to ancillary services load curves are computed using a software tool developed in Matlab, based on the monitored results. Several additional issues are tackled: numerical approximation of load curve, consumed energy computing through numerical integration and load curve forecasting. To achieve this goal, a software tool in Delphi environment has been developed. The results are discussed and the appropriate conclusions are highlighted.