Selection of Data for Modeling the Development of the Power System Using a Recurrent Artificial Neural Network

Abstract

The National Power System in Poland is now already an intelligent system, and the changes increasingly have a hybrid character, that is, structural-parameter in nature. The uniform National Power System is increasingly being supplied with electricity not only from conventional sources but also from renewable sources: solar, wind, and hydro. In addition to producers and consumers of electric power and energy, prosumers are also appearing. The power system in this diversity is equipped with various types of automated devices, which make it increasingly intelligent. In Poland, the total installed capacity of all electricity sources, both conventional and renewable, reached 72.9 GW in April 2025, of which as much as 34.8 GWcomes from renewable sources (over 47.7%), while the rapidly growing number of prosumers makes the NPS system increasingly complex at the system level. The growing complexity and diversity of the National Power System means that controlling the development of the system intelligently involves the necessity of building development models. Thus, there exists a specific research gap between the need to obtain development models of the National Power System for the purpose of development control and the possibilities of obtaining such models using artificial intelligence methods, such as neural methods. In line with research within this research gap, this study attempted to obtain a development model of the National Power System using annual statistical data on the operation of the National Power System from 1990 to 2024. This formulation of the research problem involves, among other things, the selection of data for research experiments in the area of modeling the development of the National Power System, as well as the selection of architecture and learning methods for ANNs. In the present publication, in particular, the research methodology for modeling the development of the National Power System and the obtained research results using Recurrent ANN for modeling are presented.