The present stage of development of power systems is characterized by the need to increase their intelligence, that is, the formation of an environment capable of solving the problems of effective management of all components included in the power system. One of the features of modern energy systems is an active consumer, that is, a consumer who has the opportunity to optimize the schedule of loading their capacities, planning energy consumption, intellectualization of energy consumption. In addition to the fact that there are digital systems for collecting, processing and presenting information, as well as transmitting this information via the Internet, to solve the most difficult task of managing power systems, it is necessary to develop a large number of models adequate to real processes that can describe the processes taking place in the power system under consideration. The purpose is to develop models capable of adequately describing the interrelated processes occurring in the considered elements of the power system, as well as in the aggregate of related elements; development of methods that allow solving problems of effective control of elements of energy systems on the basis of the proposed intelligent models. To solve the problem of increasing the intelligence of the power system, nonlinear adaptive models of electrical complexes of an energy-intensive enterprise included in the power system were developed on the basis of balance equations and artificial intelligence methods. The production of alumina with the properties of inertia, nonlinearity and closure is chosen as the production under consideration. This type of production was chosen as the most difficult to predict due to the unpredictability of changes in energy consumption during the transition to other modes of its operation, associated, for example, with the receipt of raw materials with a different chemical composition for processing. Static and dynamic models of energy consumption of electrotechnical complexes with continuous inertial nonlinear closed production based on artificial intelligence methods have been developed, which makes it possible to eliminate the need to use long time series of retrospective data. For the considered type of continuous energy-intensive production, dynamic characteristics of energy consumption and technological processes are obtained from the main controlled parameters, which make it possible to predict emergency situations and partly man-made disasters. On the basis of a genetic algorithm, a multiparametric optimization problem of energy management of continuous production has been solved, taking into account technological parameters and cost factors.