Renewal of the regression model for normalization of specific energy consumption

Abstract

Purpose. To develop a method of updating the regression model for the normalization of specific energy consumption in the presence of frequent and significant changes in the energy efficiency of the production process. Methodology. Analysis of existing methods of updating regression models, comparison of their possibilities, and synthesis of the method of updating the model in conditions of frequent and significant changes in the energy efficiency of the production process. Findings. It was established that in the presence of a significant number of possible variants of structural and mode changes in the energy consumption of the control object, the introduction of associated variables into the regression model is problematic, as it requires an increase in the number of experimental data in conditions of their expected heterogeneity. The flaw of the well-known regression model for normalizing the power consumption of the object of control is revealed, which consists of the fact that the model does not take into account the last values in the sequence of their appearance of experimental data obtained in the process of energy efficiency control. This reduces the accuracy of predicted energy consumption values. It is proposed to update the regression model every time after performing the energy efficiency control and sample adjustment. Adjustments are implemented by checking the homogeneity of the obtained experimental data, followed by their addition to the elements of the existing sample and removal (if necessary) from the sample of outdated data. The defined sequence of adjustment of the initial data allows timely updating of the model and implementation of the forecast of specific energy consumption, entering data reflecting the latest changes that occurred in the facility's energy supply. The proposed method of updating the model implements the approximation in time of the moment of energy efficiency control to the moment of obtaining experimental data for building a regression dependence for normalizing energy consumption values. This helps to increase the accuracy of the forecast of normalized values. A significant change in the conditions of production of products with a violation of the homogeneity of data is accompanied by a transition to the transitional mode of adjustment, where it is proposed to reduce the number of elements of the existing sample, ensuring the sequential removal of the elements furthest from the next moment of control. Extraction continues until data homogeneity is achieved. During the daily control of the efficiency of electricity consumption, the change in the values of the regression model coefficients in the process of its renewal reflects the changes in the object's electricity consumption that occurred over the last day. This allows you to separate their impact from the impact of changes that occurred earlier and to assess the level of this impact. Originality. For the first time, the shortcomings of the existing methods of updating regression models in the conditions of frequent and significant changes in the energy efficiency of the production process were identified. A method of updating the model under these conditions has been developed, which involves adjusting the sample of experimental data by changing the number of its elements and checking the homogeneity of the data. Practical value is that the sequence of actions during the implementation of the developed method of updating the regression model is defined, which allows for an increase in the accuracy of calculating the normalized values of specific energy consumption.