Identification of the Nonlinear Model proposed by the MIT for Power Transformer applying Genetic Algorithms

This paper present a technique based on Genetic Algorithms for nonlinear model parameters estimation proposed by the MIT (Massachusetts Institute of Technology) for top oil temperature prediction in power transformers that is being used in an on-line monitoring and diagnosis system installed in an 100 MVA autotransformer of Barquisimeto Substation ENELBAR Venezuela since 2003. The results of the parameters estimation by genetic algorithms are compared with previous results obtained by the parameters estimation made with Linear Minimum Square and the real measurements of the top oil temperature. Results are discussed and this model is proposed for top oil temperature prediction and diagnosis tool.