When the new model of the electricity sector began to be used, the forecast of consumption and billing of electricity began to have great importance for the electricity distributors, improving their assertiveness as companies can develop their programs for energy planning, maintenance and expansion of its distribution system and particularly to avoid fines for not spending all of his service to the consumer market. For this prediction is made it is necessary to obtain historical data of energy consumption, the load supplied by the distributor, but as these data sets are multivariate, ie, they are time series which depend on other exogenous variables took into account also the series weather (wind chill), economic (financial ratios) and the factor of time (days, holidays, weekends .....). This study proposes an alternative method to forecast sales and consumption of electricity, 15 steps forward, by developing an intelligent system, called SIPEE, MLP neural networks based on multistep and was developed for an Electricity Distributor serving much of the consumer market in the State of Rio de Janeiro, Light Electrical Services SA.