In this paper, the development and implementation of a model for forecasting spot price of the Colombian electricity market using neuronal networks is presented. With information obtained from the data base of the Colombian Wholesale Energy Market (MEM), a neuronal network for each one of the generation agents that have more influence on the market was designed; these networks were trained with the values of year 2002 and they were validated with the data of year 2003, obtaining results on the average around 7% of accuracy. The neuronal networks method, used to model each one of the generation agents, has not been used to implement models of this type in scientific researches, according to the bibliographical revision.