Design of a system for load forecasting and optimal control over enterprise energy sources under unstable power supply

This study investigates the process underlying the autonomy of industrial enterprises that face critical instability of centralized power supply and dynamic market tariffs. Given the significant losses of generating capacities and frequent interruptions in the supply of electricity, enterprises often use autonomous power sources. Optimal utilization of such sources requires economic substantiation and construction of appropriate models. The task of these models is to improve the economic efficiency and energy independence of an enterprise by automating optimal control over energy sources. This work proposes a two-level system. The first level is responsible for forecasting energy consumption. The second one is a deterministic optimization algorithm for automatic selection of a power source and economically justified control over the operational schedule of autonomous power sources. When forecasting energy consumption, two models built on the basis of Random Forest and LightGBM were compared. The models yielded average absolute errors, as a percentage of the average, of 5–7% and 8–10%, which indicates their applicability for further decision-making. Analysis of the optimization algorithm on real data revealed overall energy cost savings of 9–12% compared to unoptimized use of electricity from the grid. These results were achieved through timely switching between central and alternative power sources. Switching occurs when the use of the source becomes more economically advantageous, subject to technical constraints. The resulting system could be used by enterprises that require uninterrupted power supply and exploit generators as alternative power sources.