For energy-intensive enterprises, which include metallurgical ones, reliable prediction of electric power consumption is an urgent problem. Metallurgical production operates under the impact of a significant number of random factors, therefore it makes sense to predict power consumption within the framework of probability categories. Statistical methods make it possible to take into account systematically accumulated metering data as a massive experiment and, on the basis of such data, to establish the patterns and correlations necessary to predict energy expenditures.The specific dynamics of energy and technological indices for individual plants and the complex form of their mathematical relationships make it necessary to use a differentiated approach to formation of predictive models, where the quantitative characteristics of interrelations of energy and technological indices from the past are extrapolated to the future. Besides technological factors, the time factor is introduced to the multivariate model, which makes it possible to consider everything related to the trend of development of the object under investigation. But in this case, the problem of constructing a multivariate predictive model becomes much more complex, since the dynamic series under investigation may contain nonlinear trends in the form of second-order or higher polynomials, which makes it hard to use them for practical purposes.
展开▼
