Control of Electricity Consumption in an Enterprise Using Reinforcement Learning

The paper considers the problems of unawareness in hierarchical management of electricity consumption and the use of artificial intelligence, including asymmetric awareness of electricity consumers. The issues of integrating machine learning and control theory in monitoring electricity consumption are analyzed. A three-level structure of electricity consumption at an enterprise is modelled, with the director at the top level, the manager at the middle level, and the worker at the bottom. A mechanism for controlling the manager's electricity consumption has been developed, in which the director uses supervised learning. The manager's reactions and the director's optimal policy are determined. A mechanism for controlling the worker's electricity consumption has also been developed, in which the manager uses reinforcement learning. The worker's reactions and the manager's optimal policy are determined. The combination of these mechanisms forms control mechanism for electricity consumption at an enterprise.