In this paper we consider the problem of effective forecasting of electricity consumption for multiple objects in automatic mode. We propose an approach to automation of forecasting process, based on CRISP-DM. There were considered methods of computational intelligence for the data preprocessing - filling the gaps and identification of the emission, methods of construction and adjustment of forecasting models. As the computational basis are used connective models, based on neural networks and on constructive neural networks. There was shown the system architecture that implements the mechanisms of computational intelligence, and was presented the results of the tests.
Keywords: consumption of electricity, forecasting automation, computational intelligence, identification of the emission, connective models.