• DocumentCode
    3242332
  • Title

    Comparison study of an in-house developed tool and commercial software for ANN modeling

  • Author

    Smrekar, J. ; Assadi, M.

  • Author_Institution
    Dept. of Mech. & Struct. Eng. & Mater. Sci., Univ. of Stavanger, Stavanger, Norway
  • fYear
    2010
  • fDate
    29-31 Oct. 2010
  • Firstpage
    1510
  • Lastpage
    1516
  • Abstract
    This study presents a comparison of an in-house developed tool called Artificial Neural Network Model Developer (ANNMD) and commercial software for modeling with Artificial Neural Networks (ANNs). The development of ANNMD is based on initiatives from previous experiences in ANN modeling of complex energy systems which require extensive examination of measured parameters in order to obtain reliable ANN models. The ANNMD is a generic tool the main idea of which is to perform fast, user-friendly modeling and with automated sensitivity analysis (SA) in conjunction with parameters that are important in every ANN modeling. In addition, commercial software for ANN modeling - which has been present on the market for more than a decade - was used. The comparison was established on the same basis with respect to network configuration, main training settings and data used, hence enabling as objective analysis as possible. Modeling time, number of trainings and model´s accuracy obtained by the two tools were the measures for the comparison.
  • Keywords
    neural nets; power engineering computing; power systems; ANNMD; artificial neural network model developer; automated sensitivity analysis; commercial software; complex energy systems; in-house developed tool; objective analysis; training settings; user friendly modeling; Artificial neural networks; Artificial Neural Networks; Comparison study; Energy systems; Generic tool; Modeling;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Industrial Engineering and Engineering Management (IE&EM), 2010 IEEE 17Th International Conference on
  • Conference_Location
    Xiamen
  • Print_ISBN
    978-1-4244-6483-8
  • Type

    conf

  • DOI
    10.1109/ICIEEM.2010.5646037
  • Filename
    5646037