Electricity Costs Control with Supervised Learning

Models and methods of organizational control of electricity costs in a 3-level company under conditions of uncertainty are considered. The human factor is taken into account in the form of activities associated with the presence of the elements of the lower levels of the company’s hierarchy of their own goals. These goals do not coincide with the goals of senior managers. Such active elements may choose their electricity costs metrics to maximize their own target functions, rather than the target functions of senior managers. With insufficient awareness of managers, such activity of the elements of the company can lead to overestimation of its electricity costs. To avoid this, the managers of the top two levels are advised to use training procedures based on machine learning algorithms, as well as supervisor’s instructions. In addition to this, the company’s electricity cost management mechanism includes ranking and incentive procedures at the lower two levels of the company. The solutions of the games of the active elements of these levels of the company with higher-level managers are determined. These solutions depend on the used mechanism for managing the costs of electricity. Sufficient conditions have been found for the synthesis of a mechanism for minimizing these costs, in which the active elements of the company make full use of their opportunities to reduce these costs. The results obtained are illustrated by the example of a mechanism for reducing electricity costs for wagon repairs in the management system of JSC Russian Railways.